zdoom/gzdoom
0
0
Fork 0
mirror of https://github.com/ZDoom/gzdoom.git synced 2024-11-27 14:22:13 +00:00
Code Issues Projects Releases Packages Wiki Activity
gzdoom/src/dobjtype.cpp

3767 lines
101 KiB
C++
Raw Normal View History

Normalize line endings
2016-03-01 15:47:10 +00:00
/*
** dobjtype.cpp
** Implements the type information class
**
**---------------------------------------------------------------------------
** Copyright 1998-2010 Randy Heit
** All rights reserved.
**
** Redistribution and use in source and binary forms, with or without
** modification, are permitted provided that the following conditions
** are met:
**
** 1. Redistributions of source code must retain the above copyright
** notice, this list of conditions and the following disclaimer.
** 2. Redistributions in binary form must reproduce the above copyright
** notice, this list of conditions and the following disclaimer in the
** documentation and/or other materials provided with the distribution.
** 3. The name of the author may not be used to endorse or promote products
** derived from this software without specific prior written permission.
**
** THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
** IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
** OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
** IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
** INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
** NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
** DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
** THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
** THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**---------------------------------------------------------------------------
**
*/
// HEADER FILES ------------------------------------------------------------
#include <float.h>
#include <limits>
#include "dobject.h"
#include "i_system.h"
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
#include "serializer.h"
Normalize line endings
2016-03-01 15:47:10 +00:00
#include "actor.h"
#include "templates.h"
#include "autosegs.h"
#include "v_text.h"
#include "a_pickups.h"
#include "d_player.h"
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
#include "doomerrors.h"
Normalize line endings
2016-03-01 15:47:10 +00:00
#include "fragglescript/t_fs.h"
- split up zscript/shared/inventory.txt. - moved health items to their own file. - scriptified ScoreItem and MapRevealer whose entire functionality was a small TryPickup method. - fixed: bit fields in global variables were not correctly written. This should conclude the inventory cleanup. It is now possible again to find things in there.
2016-11-30 12:36:13 +00:00
#include "a_keys.h"
Normalize line endings
2016-03-01 15:47:10 +00:00
// MACROS ------------------------------------------------------------------
// TYPES -------------------------------------------------------------------
// EXTERNAL FUNCTION PROTOTYPES --------------------------------------------
// PUBLIC FUNCTION PROTOTYPES ----------------------------------------------
// PRIVATE FUNCTION PROTOTYPES ---------------------------------------------
// EXTERNAL DATA DECLARATIONS ----------------------------------------------
- added 'strictdecorate' CVAR. If this is set to 'true', any DECORATE error that had to be demoted to a warning for backwards compatibility will be treated as an actual error.
2016-06-08 08:56:11 +00:00
EXTERN_CVAR(Bool, strictdecorate);
Normalize line endings
2016-03-01 15:47:10 +00:00
// PUBLIC DATA DEFINITIONS -------------------------------------------------
- Turned DropItem into a plain struct again like it was before the scripting branch got merged. Making this an object had little to no advantage, except being able to remove the deleter code. Now, with some of the class data already being allocated in a memory arena so that freeing it is easier, this can also be used for the drop item lists which makes it unnecessary to subject them to the GC. This also merges the memory arenas for VM functions and flat pointers because both get deleted at the same time so they can share the same one.
2017-02-08 19:37:22 +00:00
FMemArena ClassDataAllocator(32768); // use this for all static class data that can be released in bulk when the type system is shut down.
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
Normalize line endings
2016-03-01 15:47:10 +00:00
FTypeTable TypeTable;
TArray<PClass *> PClass::AllClasses;
- fixed: the restart CCMD must clear out all statically stored pointers to VM functions because they do not survive the shutdown that is needed before loading new data.
2017-01-31 12:41:23 +00:00
TArray<VMFunction**> PClass::FunctionPtrList;
Normalize line endings
2016-03-01 15:47:10 +00:00
bool PClass::bShutdown;
- scriptified the sector actions.
2017-01-13 12:51:47 +00:00
bool PClass::bVMOperational;
Normalize line endings
2016-03-01 15:47:10 +00:00
PErrorType *TypeError;
- added automatic type deduction capabilities for local variables. If you type 'let variable = value;' the type of 'variable' will be deduced from the given value. This is mostly of interest for type casting pointers, because 'let p = Classtype(objectvar);' does not require writing the class type name twice.
2016-12-05 12:24:42 +00:00
PErrorType *TypeAuto;
Normalize line endings
2016-03-01 15:47:10 +00:00
PVoidType *TypeVoid;
PInt *TypeSInt8, *TypeUInt8;
PInt *TypeSInt16, *TypeUInt16;
PInt *TypeSInt32, *TypeUInt32;
PBool *TypeBool;
PFloat *TypeFloat32, *TypeFloat64;
PString *TypeString;
PName *TypeName;
PSound *TypeSound;
PColor *TypeColor;
- added two new integral types SpriteID and TextureID, both are needed to allow proper serialization as they require something different to be written out than a regular integer.
2016-11-21 12:45:33 +00:00
PTextureID *TypeTextureID;
PSpriteID *TypeSpriteID;
Normalize line endings
2016-03-01 15:47:10 +00:00
PStatePointer *TypeState;
- streamlined font handling for scripts a bit. - moved the two 'you raised the alarm' messages for Strife to the string table
2017-02-05 12:14:22 +00:00
PPointer *TypeFont;
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
PStateLabel *TypeStateLabel;
- parse the list of class/struct fields and add the PField entries to their type. - fixed a few issues with trying to access a class's Type before it got initialized.
2016-10-09 13:47:31 +00:00
PStruct *TypeVector2;
PStruct *TypeVector3;
- added per-channel access for color variables. However, since they are locally stored in registers, it required a minor bit of fudging for this case. - make sure that a PFunction's implementation is always initialized before starting the code generator.
2016-11-21 11:38:39 +00:00
PStruct *TypeColorStruct;
- scriptified Strife's coins. - added a String class to allow attaching methods to the builtin string type. This works by checking if the left side of the member accessor is a string and just replacing the tyoe in this one place, all the rest is automatic.
2016-11-28 17:15:18 +00:00
PStruct *TypeStringStruct;
- preparations for null pointer support. - add pointer support to FxAssign.
2016-10-21 17:18:39 +00:00
PPointer *TypeNullPtr;
some groundwork for the implementation of dynamic arrays - created script exports for all relevant functions with all integral types. - created script side definitions for the underlying data types. - added a void pointer type so that the prototype for the pointer array can use a generic type every pointer can be assigned to.
2017-01-10 23:57:31 +00:00
PPointer *TypeVoidPtr;
Normalize line endings
2016-03-01 15:47:10 +00:00
// PRIVATE DATA DEFINITIONS ------------------------------------------------
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// A harmless non-nullptr FlatPointer for classes without pointers.
Normalize line endings
2016-03-01 15:47:10 +00:00
static const size_t TheEnd = ~(size_t)0;
// CODE --------------------------------------------------------------------
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PErrorType, false, false)
IMPLEMENT_CLASS(PVoidType, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
void DumpTypeTable()
{
int used = 0;
int min = INT_MAX;
int max = 0;
int all = 0;
int lens[10] = {0};
for (size_t i = 0; i < countof(TypeTable.TypeHash); ++i)
{
int len = 0;
Printf("%4zu:", i);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
for (PType *ty = TypeTable.TypeHash[i]; ty != nullptr; ty = ty->HashNext)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
Printf(" -> %s", ty->DescriptiveName());
Normalize line endings
2016-03-01 15:47:10 +00:00
len++;
all++;
}
if (len != 0)
{
used++;
if (len < min)
min = len;
if (len > max)
max = len;
}
if (len < (int)countof(lens))
{
lens[len]++;
}
Printf("\n");
}
Printf("Used buckets: %d/%lu (%.2f%%) for %d entries\n", used, countof(TypeTable.TypeHash), double(used)/countof(TypeTable.TypeHash)*100, all);
Printf("Min bucket size: %d\n", min);
Printf("Max bucket size: %d\n", max);
Printf("Avg bucket size: %.2f\n", double(all) / used);
int j,k;
for (k = countof(lens)-1; k > 0; --k)
if (lens[k])
break;
for (j = 0; j <= k; ++j)
Printf("Buckets of len %d: %d (%.2f%%)\n", j, lens[j], j!=0?double(lens[j])/used*100:-1.0);
}
/* PType ******************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PType, true, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PType Parameterized Constructor
//
//==========================================================================
PType::PType(unsigned int size, unsigned int align)
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: Size(size), Align(align), HashNext(nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Type";
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
loadOp = OP_NOP;
storeOp = OP_NOP;
moveOp = OP_NOP;
RegType = REGT_NIL;
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
RegCount = 1;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PType Destructor
//
//==========================================================================
PType::~PType()
{
}
//==========================================================================
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//
// PType :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PType::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
assert(0 && "Cannot write value for this type");
}
//==========================================================================
//
// PType :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PType::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
assert(0 && "Cannot read value for this type");
return false;
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PType :: SetDefaultValue
//
//==========================================================================
void PType::SetDefaultValue(void *base, unsigned offset, TArray<FTypeAndOffset> *stroffs) const
{
}
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
//==========================================================================
//
// PType :: SetDefaultValue
//
//==========================================================================
void PType::SetPointer(void *base, unsigned offset, TArray<size_t> *stroffs) const
{
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
void PType::SetPointerArray(void *base, unsigned offset, TArray<size_t> *stroffs) const
{
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PType :: InitializeValue
//
//==========================================================================
void PType::InitializeValue(void *addr, const void *def) const
{
}
//==========================================================================
//
// PType :: DestroyValue
//
//==========================================================================
void PType::DestroyValue(void *addr) const
{
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PType :: SetValue
//
//==========================================================================
void PType::SetValue(void *addr, int val)
{
Add double variants of SetValue() for numeric PTypes
2016-03-30 03:05:25 +00:00
assert(0 && "Cannot set int value for this type");
}
void PType::SetValue(void *addr, double val)
{
assert(0 && "Cannot set float value for this type");
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PType :: GetValue
//
//==========================================================================
int PType::GetValueInt(void *addr) const
{
assert(0 && "Cannot get value for this type");
return 0;
}
Add GetValueFloat() for numeric PTypes
2016-03-30 03:24:59 +00:00
double PType::GetValueFloat(void *addr) const
{
assert(0 && "Cannot get value for this type");
return 0;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PType :: IsMatch
//
//==========================================================================
bool PType::IsMatch(intptr_t id1, intptr_t id2) const
{
return false;
}
//==========================================================================
//
// PType :: GetTypeIDs
//
//==========================================================================
void PType::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = 0;
id2 = 0;
}
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
//==========================================================================
//
// PType :: GetTypeIDs
//
//==========================================================================
const char *PType::DescriptiveName() const
{
return mDescriptiveName.GetChars();
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PType :: StaticInit STATIC
//
//==========================================================================
void PType::StaticInit()
{
// Create types and add them type the type table.
TypeTable.AddType(TypeError = new PErrorType);
- added automatic type deduction capabilities for local variables. If you type 'let variable = value;' the type of 'variable' will be deduced from the given value. This is mostly of interest for type casting pointers, because 'let p = Classtype(objectvar);' does not require writing the class type name twice.
2016-12-05 12:24:42 +00:00
TypeTable.AddType(TypeAuto = new PErrorType(2));
Normalize line endings
2016-03-01 15:47:10 +00:00
TypeTable.AddType(TypeVoid = new PVoidType);
TypeTable.AddType(TypeSInt8 = new PInt(1, false));
TypeTable.AddType(TypeUInt8 = new PInt(1, true));
TypeTable.AddType(TypeSInt16 = new PInt(2, false));
TypeTable.AddType(TypeUInt16 = new PInt(2, true));
TypeTable.AddType(TypeSInt32 = new PInt(4, false));
TypeTable.AddType(TypeUInt32 = new PInt(4, true));
TypeTable.AddType(TypeBool = new PBool);
TypeTable.AddType(TypeFloat32 = new PFloat(4));
TypeTable.AddType(TypeFloat64 = new PFloat(8));
TypeTable.AddType(TypeString = new PString);
TypeTable.AddType(TypeName = new PName);
TypeTable.AddType(TypeSound = new PSound);
TypeTable.AddType(TypeColor = new PColor);
TypeTable.AddType(TypeState = new PStatePointer);
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
TypeTable.AddType(TypeStateLabel = new PStateLabel);
- preparations for null pointer support. - add pointer support to FxAssign.
2016-10-21 17:18:39 +00:00
TypeTable.AddType(TypeNullPtr = new PPointer);
- added two new integral types SpriteID and TextureID, both are needed to allow proper serialization as they require something different to be written out than a regular integer.
2016-11-21 12:45:33 +00:00
TypeTable.AddType(TypeSpriteID = new PSpriteID);
TypeTable.AddType(TypeTextureID = new PTextureID);
Normalize line endings
2016-03-01 15:47:10 +00:00
some groundwork for the implementation of dynamic arrays - created script exports for all relevant functions with all integral types. - created script side definitions for the underlying data types. - added a void pointer type so that the prototype for the pointer array can use a generic type every pointer can be assigned to.
2017-01-10 23:57:31 +00:00
TypeVoidPtr = NewPointer(TypeVoid, false);
- scriptified Strife's coins. - added a String class to allow attaching methods to the builtin string type. This works by checking if the left side of the member accessor is a string and just replacing the tyoe in this one place, all the rest is automatic.
2016-11-28 17:15:18 +00:00
TypeColorStruct = NewStruct("@ColorStruct", nullptr); //This name is intentionally obfuscated so that it cannot be used explicitly. The point of this type is to gain access to the single channels of a color value.
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
TypeStringStruct = NewNativeStruct("Stringstruct", nullptr);
- streamlined font handling for scripts a bit. - moved the two 'you raised the alarm' messages for Strife to the string table
2017-02-05 12:14:22 +00:00
TypeFont = NewPointer(NewNativeStruct("Font", nullptr));
- added per-channel access for color variables. However, since they are locally stored in registers, it required a minor bit of fudging for this case. - make sure that a PFunction's implementation is always initialized before starting the code generator.
2016-11-21 11:38:39 +00:00
#ifdef __BIG_ENDIAN__
TypeColorStruct->AddField(NAME_a, TypeUInt8);
TypeColorStruct->AddField(NAME_r, TypeUInt8);
TypeColorStruct->AddField(NAME_g, TypeUInt8);
TypeColorStruct->AddField(NAME_b, TypeUInt8);
#else
TypeColorStruct->AddField(NAME_b, TypeUInt8);
TypeColorStruct->AddField(NAME_g, TypeUInt8);
TypeColorStruct->AddField(NAME_r, TypeUInt8);
TypeColorStruct->AddField(NAME_a, TypeUInt8);
#endif
- parse the list of class/struct fields and add the PField entries to their type. - fixed a few issues with trying to access a class's Type before it got initialized.
2016-10-09 13:47:31 +00:00
TypeVector2 = new PStruct(NAME_Vector2, nullptr);
TypeVector2->AddField(NAME_X, TypeFloat64);
TypeVector2->AddField(NAME_Y, TypeFloat64);
TypeTable.AddType(TypeVector2);
- fixed GetOps() call and added info for vectors.
2016-10-28 11:30:19 +00:00
TypeVector2->loadOp = OP_LV2;
TypeVector2->storeOp = OP_SV2;
TypeVector2->moveOp = OP_MOVEV2;
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
TypeVector2->RegType = REGT_FLOAT;
TypeVector2->RegCount = 2;
- parse the list of class/struct fields and add the PField entries to their type. - fixed a few issues with trying to access a class's Type before it got initialized.
2016-10-09 13:47:31 +00:00
TypeVector3 = new PStruct(NAME_Vector3, nullptr);
TypeVector3->AddField(NAME_X, TypeFloat64);
TypeVector3->AddField(NAME_Y, TypeFloat64);
TypeVector3->AddField(NAME_Z, TypeFloat64);
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
// allow accessing xy as a vector2. This is not supposed to be serialized so it's marked transient
TypeVector3->Symbols.AddSymbol(new PField(NAME_XY, TypeVector2, VARF_Transient, 0));
- parse the list of class/struct fields and add the PField entries to their type. - fixed a few issues with trying to access a class's Type before it got initialized.
2016-10-09 13:47:31 +00:00
TypeTable.AddType(TypeVector3);
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
TypeVector3->loadOp = OP_LV3;
TypeVector3->storeOp = OP_SV3;
TypeVector3->moveOp = OP_MOVEV3;
TypeVector3->RegType = REGT_FLOAT;
TypeVector3->RegCount = 3;
- parse the list of class/struct fields and add the PField entries to their type. - fixed a few issues with trying to access a class's Type before it got initialized.
2016-10-09 13:47:31 +00:00
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_sByte, TypeSInt8));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Byte, TypeUInt8));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Short, TypeSInt16));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_uShort, TypeUInt16));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Int, TypeSInt32));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_uInt, TypeUInt32));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Bool, TypeBool));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Float, TypeFloat64));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Double, TypeFloat64));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Float32, TypeFloat32));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Float64, TypeFloat64));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_String, TypeString));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Name, TypeName));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Sound, TypeSound));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Color, TypeColor));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_State, TypeState));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Vector2, TypeVector2));
Namespaces.GlobalNamespace->Symbols.AddSymbol(new PSymbolType(NAME_Vector3, TypeVector3));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
/* PBasicType *************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PBasicType, true, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PBasicType Default Constructor
//
//==========================================================================
PBasicType::PBasicType()
{
}
//==========================================================================
//
// PBasicType Parameterized Constructor
//
//==========================================================================
PBasicType::PBasicType(unsigned int size, unsigned int align)
: PType(size, align)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "BasicType";
Normalize line endings
2016-03-01 15:47:10 +00:00
}
/* PCompoundType **********************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PCompoundType, true, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PNamedType *************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PNamedType, true, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PNamedType :: IsMatch
//
//==========================================================================
bool PNamedType::IsMatch(intptr_t id1, intptr_t id2) const
{
const DObject *outer = (const DObject *)id1;
FName name = (ENamedName)(intptr_t)id2;
return Outer == outer && TypeName == name;
}
//==========================================================================
//
// PNamedType :: GetTypeIDs
//
//==========================================================================
void PNamedType::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)Outer;
id2 = TypeName;
}
/* PInt *******************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PInt, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PInt Default Constructor
//
//==========================================================================
PInt::PInt()
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
: PBasicType(4, 4), Unsigned(false), IntCompatible(true)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "SInt32";
Normalize line endings
2016-03-01 15:47:10 +00:00
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Min, this, -0x7FFFFFFF - 1));
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Max, this, 0x7FFFFFFF));
- fixed GetOps() call and added info for vectors.
2016-10-28 11:30:19 +00:00
SetOps();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PInt Parameterized Constructor
//
//==========================================================================
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
PInt::PInt(unsigned int size, bool unsign, bool compatible)
: PBasicType(size, size), Unsigned(unsign), IntCompatible(compatible)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("%cInt%d", unsign? 'U':'S', size);
- restrict 8 and 16 bit integer types to struct and class members - and I am not really sure if they should even be allowed there, aside from legacy readonly properties. For 32 bit floats the same would hold, but those are already mapped to 64 bit doubles silently. ZDoom only uses these types in a very few isolated places, and even those can be removed without problems, so it's very doubtful that having support for these types is of any benefit - on the other hand, having them will most likely introduce more code than is saved in the data by using them...
2016-10-16 10:47:26 +00:00
MemberOnly = (size < 4);
Normalize line endings
2016-03-01 15:47:10 +00:00
if (!unsign)
{
Correct two overflows in PInt ctor
2017-02-22 23:58:08 +00:00
int maxval = (1u << ((8 * size) - 1)) - 1; // compute as unsigned to prevent overflow before -1
Normalize line endings
2016-03-01 15:47:10 +00:00
int minval = -maxval - 1;
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Min, this, minval));
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Max, this, maxval));
}
else
{
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Min, this, 0u));
Correct two overflows in PInt ctor
2017-02-22 23:58:08 +00:00
Symbols.AddSymbol(new PSymbolConstNumeric(NAME_Max, this, (1u << ((8 * size) - 1))));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
SetOps();
}
void PInt::SetOps()
{
moveOp = OP_MOVE;
RegType = REGT_INT;
if (Size == 4)
{
storeOp = OP_SW;
loadOp = OP_LW;
}
else if (Size == 1)
{
storeOp = OP_SB;
loadOp = Unsigned ? OP_LBU : OP_LB;
}
else if (Size == 2)
{
storeOp = OP_SH;
loadOp = Unsigned ? OP_LHU : OP_LH;
}
else
{
assert(0 && "Unhandled integer size");
storeOp = OP_NOP;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PInt :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PInt::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (Size == 8 && Unsigned)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
// this is a special case that cannot be represented by an int64_t.
uint64_t val = *(uint64_t*)addr;
ar(key, val);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
else
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
int64_t val;
switch (Size)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
case 1:
val = Unsigned ? *(uint8_t*)addr : *(int8_t*)addr;
break;
case 2:
val = Unsigned ? *(uint16_t*)addr : *(int16_t*)addr;
break;
case 4:
val = Unsigned ? *(uint32_t*)addr : *(int32_t*)addr;
break;
case 8:
val = *(int64_t*)addr;
break;
default:
return; // something invalid
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, val);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
//==========================================================================
//
// PInt :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PInt::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
NumericValue val;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, val);
if (val.type == NumericValue::NM_invalid) return false; // not found or usable
if (val.type == NumericValue::NM_float) val.signedval = (int64_t)val.floatval;
// No need to check the unsigned state here. Downcasting to smaller types will yield the same result for both.
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
switch (Size)
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
case 1:
*(uint8_t*)addr = (uint8_t)val.signedval;
break;
case 2:
*(uint16_t*)addr = (uint16_t)val.signedval;
break;
case 4:
*(uint32_t*)addr = (uint32_t)val.signedval;
break;
case 8:
*(uint64_t*)addr = (uint64_t)val.signedval;
break;
default:
return false; // something invalid
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
return true;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PInt :: SetValue
//
//==========================================================================
void PInt::SetValue(void *addr, int val)
{
assert(((intptr_t)addr & (Align - 1)) == 0 && "unaligned address");
if (Size == 4)
{
*(int *)addr = val;
}
else if (Size == 1)
{
*(BYTE *)addr = val;
}
else if (Size == 2)
{
*(WORD *)addr = val;
}
else if (Size == 8)
{
*(QWORD *)addr = val;
}
else
{
assert(0 && "Unhandled integer size");
}
}
Add double variants of SetValue() for numeric PTypes
2016-03-30 03:05:25 +00:00
void PInt::SetValue(void *addr, double val)
{
SetValue(addr, (int)val);
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PInt :: GetValueInt
//
//==========================================================================
int PInt::GetValueInt(void *addr) const
{
assert(((intptr_t)addr & (Align - 1)) == 0 && "unaligned address");
if (Size == 4)
{
return *(int *)addr;
}
else if (Size == 1)
{
return Unsigned ? *(BYTE *)addr : *(SBYTE *)addr;
}
else if (Size == 2)
{
return Unsigned ? *(WORD *)addr : *(SWORD *)addr;
}
else if (Size == 8)
{ // truncated output
return (int)*(QWORD *)addr;
}
else
{
assert(0 && "Unhandled integer size");
return 0;
}
}
Add GetValueFloat() for numeric PTypes
2016-03-30 03:24:59 +00:00
//==========================================================================
//
// PInt :: GetValueFloat
//
//==========================================================================
double PInt::GetValueFloat(void *addr) const
{
return GetValueInt(addr);
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PInt :: GetStoreOp
//
//==========================================================================
/* PBool ******************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PBool, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PBool Default Constructor
//
//==========================================================================
PBool::PBool()
: PInt(sizeof(bool), true)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Bool";
MemberOnly = false;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
/* PFloat *****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PFloat, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PFloat Default Constructor
//
//==========================================================================
PFloat::PFloat()
: PBasicType(8, 8)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Float";
Normalize line endings
2016-03-01 15:47:10 +00:00
SetDoubleSymbols();
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
SetOps();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PFloat Parameterized Constructor
//
//==========================================================================
PFloat::PFloat(unsigned int size)
: PBasicType(size, size)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("Float%d", size);
Normalize line endings
2016-03-01 15:47:10 +00:00
if (size == 8)
{
SetDoubleSymbols();
}
else
{
assert(size == 4);
- restrict 8 and 16 bit integer types to struct and class members - and I am not really sure if they should even be allowed there, aside from legacy readonly properties. For 32 bit floats the same would hold, but those are already mapped to 64 bit doubles silently. ZDoom only uses these types in a very few isolated places, and even those can be removed without problems, so it's very doubtful that having support for these types is of any benefit - on the other hand, having them will most likely introduce more code than is saved in the data by using them...
2016-10-16 10:47:26 +00:00
MemberOnly = true;
Normalize line endings
2016-03-01 15:47:10 +00:00
SetSingleSymbols();
}
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
SetOps();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PFloat :: SetDoubleSymbols
//
// Setup constant values for 64-bit floats.
//
//==========================================================================
void PFloat::SetDoubleSymbols()
{
static const SymbolInitF symf[] =
{
{ NAME_Min_Normal, DBL_MIN },
{ NAME_Max, DBL_MAX },
{ NAME_Epsilon, DBL_EPSILON },
{ NAME_NaN, std::numeric_limits<double>::quiet_NaN() },
{ NAME_Infinity, std::numeric_limits<double>::infinity() },
{ NAME_Min_Denormal, std::numeric_limits<double>::denorm_min() }
};
static const SymbolInitI symi[] =
{
{ NAME_Dig, DBL_DIG },
{ NAME_Min_Exp, DBL_MIN_EXP },
{ NAME_Max_Exp, DBL_MAX_EXP },
{ NAME_Mant_Dig, DBL_MANT_DIG },
{ NAME_Min_10_Exp, DBL_MIN_10_EXP },
{ NAME_Max_10_Exp, DBL_MAX_10_EXP }
};
SetSymbols(symf, countof(symf));
SetSymbols(symi, countof(symi));
}
//==========================================================================
//
// PFloat :: SetSingleSymbols
//
// Setup constant values for 32-bit floats.
//
//==========================================================================
void PFloat::SetSingleSymbols()
{
static const SymbolInitF symf[] =
{
{ NAME_Min_Normal, FLT_MIN },
{ NAME_Max, FLT_MAX },
{ NAME_Epsilon, FLT_EPSILON },
{ NAME_NaN, std::numeric_limits<float>::quiet_NaN() },
{ NAME_Infinity, std::numeric_limits<float>::infinity() },
{ NAME_Min_Denormal, std::numeric_limits<float>::denorm_min() }
};
static const SymbolInitI symi[] =
{
{ NAME_Dig, FLT_DIG },
{ NAME_Min_Exp, FLT_MIN_EXP },
{ NAME_Max_Exp, FLT_MAX_EXP },
{ NAME_Mant_Dig, FLT_MANT_DIG },
{ NAME_Min_10_Exp, FLT_MIN_10_EXP },
{ NAME_Max_10_Exp, FLT_MAX_10_EXP }
};
SetSymbols(symf, countof(symf));
SetSymbols(symi, countof(symi));
}
//==========================================================================
//
// PFloat :: SetSymbols
//
//==========================================================================
void PFloat::SetSymbols(const PFloat::SymbolInitF *sym, size_t count)
{
for (size_t i = 0; i < count; ++i)
{
Symbols.AddSymbol(new PSymbolConstNumeric(sym[i].Name, this, sym[i].Value));
}
}
void PFloat::SetSymbols(const PFloat::SymbolInitI *sym, size_t count)
{
for (size_t i = 0; i < count; ++i)
{
Symbols.AddSymbol(new PSymbolConstNumeric(sym[i].Name, this, sym[i].Value));
}
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PFloat :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PFloat::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
if (Size == 8)
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, *(double*)addr);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, *(float*)addr);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
//==========================================================================
//
// PFloat :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PFloat::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
NumericValue val;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, val);
if (val.type == NumericValue::NM_invalid) return false; // not found or usable
else if (val.type == NumericValue::NM_signed) val.floatval = (double)val.signedval;
else if (val.type == NumericValue::NM_unsigned) val.floatval = (double)val.unsignedval;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (Size == 8)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
*(double*)addr = val.floatval;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
else
{
*(float*)addr = (float)val.floatval;
}
return true;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PFloat :: SetValue
//
//==========================================================================
void PFloat::SetValue(void *addr, int val)
Add double variants of SetValue() for numeric PTypes
2016-03-30 03:05:25 +00:00
{
return SetValue(addr, (double)val);
}
void PFloat::SetValue(void *addr, double val)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
assert(((intptr_t)addr & (Align - 1)) == 0 && "unaligned address");
if (Size == 4)
{
*(float *)addr = (float)val;
}
else
{
assert(Size == 8);
*(double *)addr = val;
}
}
//==========================================================================
//
// PFloat :: GetValueInt
//
//==========================================================================
int PFloat::GetValueInt(void *addr) const
Add GetValueFloat() for numeric PTypes
2016-03-30 03:24:59 +00:00
{
return xs_ToInt(GetValueFloat(addr));
}
//==========================================================================
//
// PFloat :: GetValueFloat
//
//==========================================================================
double PFloat::GetValueFloat(void *addr) const
Normalize line endings
2016-03-01 15:47:10 +00:00
{
assert(((intptr_t)addr & (Align - 1)) == 0 && "unaligned address");
if (Size == 4)
{
Add GetValueFloat() for numeric PTypes
2016-03-30 03:24:59 +00:00
return *(float *)addr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
else
{
assert(Size == 8);
Add GetValueFloat() for numeric PTypes
2016-03-30 03:24:59 +00:00
return *(double *)addr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
}
//==========================================================================
//
// PFloat :: GetStoreOp
//
//==========================================================================
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
void PFloat::SetOps()
Normalize line endings
2016-03-01 15:47:10 +00:00
{
if (Size == 4)
{
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
storeOp = OP_SSP;
loadOp = OP_LSP;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
else
{
assert(Size == 8);
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
storeOp = OP_SDP;
loadOp = OP_LDP;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
moveOp = OP_MOVEF;
RegType = REGT_FLOAT;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
/* PString ****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PString, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PString Default Constructor
//
//==========================================================================
PString::PString()
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
: PBasicType(sizeof(FString), alignof(FString))
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "String";
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
storeOp = OP_SS;
loadOp = OP_LS;
moveOp = OP_MOVES;
RegType = REGT_STRING;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PString :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PString::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar(key, *(FString*)addr);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
//==========================================================================
//
// PString :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PString::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const char *cptr;
ar.StringPtr(key, cptr);
if (cptr == nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return false;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
*(FString*)addr = cptr;
return true;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PString :: SetDefaultValue
//
//==========================================================================
void PString::SetDefaultValue(void *base, unsigned offset, TArray<FTypeAndOffset> *special) const
{
- changed PClass:AddField so that it doesn't create defaults if none are present. The ZScript compiler will not create any defaults until after all fields have been added and this function created bad data if it was run on a class without defaults. - added checks for duplicate field names. - moved the tree node symbol tables out of PSymbolTreeNode to the worker data. That symbol is a bad location because it restricts the usefulness of the symbol class which is also needed for variables which use different AST structs. - fixed some memory management issues with the work classes for the compiler that became apparent after moving the symbol tables in there. In several places these were copied around, possibly losing data.
2016-10-09 23:18:47 +00:00
if (base != nullptr) new((BYTE *)base + offset) FString;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (special != nullptr)
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
special->Push(std::make_pair(this, offset));
}
}
//==========================================================================
//
// PString :: InitializeValue
//
//==========================================================================
void PString::InitializeValue(void *addr, const void *def) const
{
- changed PClass:AddField so that it doesn't create defaults if none are present. The ZScript compiler will not create any defaults until after all fields have been added and this function created bad data if it was run on a class without defaults. - added checks for duplicate field names. - moved the tree node symbol tables out of PSymbolTreeNode to the worker data. That symbol is a bad location because it restricts the usefulness of the symbol class which is also needed for variables which use different AST structs. - fixed some memory management issues with the work classes for the compiler that became apparent after moving the symbol tables in there. In several places these were copied around, possibly losing data.
2016-10-09 23:18:47 +00:00
if (def != nullptr)
{
new(addr) FString(*(FString *)def);
}
else
{
new(addr) FString;
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
}
//==========================================================================
//
// PString :: DestroyValue
//
//==========================================================================
void PString::DestroyValue(void *addr) const
{
((FString *)addr)->~FString();
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PName ******************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PName, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PName Default Constructor
//
//==========================================================================
PName::PName()
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
: PInt(sizeof(FName), true, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Name";
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
assert(sizeof(FName) == alignof(FName));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PName :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PName::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const char *cptr = ((const FName*)addr)->GetChars();
ar.StringPtr(key, cptr);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
//==========================================================================
//
// PName :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PName::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const char *cptr;
ar.StringPtr(key, cptr);
if (cptr == nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return false;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
*(FName*)addr = FName(cptr);
return true;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
- added two new integral types SpriteID and TextureID, both are needed to allow proper serialization as they require something different to be written out than a regular integer.
2016-11-21 12:45:33 +00:00
/* PSpriteID ******************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PSpriteID, false, false)
- added two new integral types SpriteID and TextureID, both are needed to allow proper serialization as they require something different to be written out than a regular integer.
2016-11-21 12:45:33 +00:00
//==========================================================================
//
// PName Default Constructor
//
//==========================================================================
PSpriteID::PSpriteID()
: PInt(sizeof(int), true, true)
{
mDescriptiveName = "SpriteID";
}
//==========================================================================
//
// PName :: WriteValue
//
//==========================================================================
void PSpriteID::WriteValue(FSerializer &ar, const char *key, const void *addr) const
{
int32_t val = *(int*)addr;
ar.Sprite(key, val, nullptr);
}
//==========================================================================
//
// PName :: ReadValue
//
//==========================================================================
bool PSpriteID::ReadValue(FSerializer &ar, const char *key, void *addr) const
{
int32_t val;
ar.Sprite(key, val, nullptr);
*(int*)addr = val;
return true;
}
/* PTextureID ******************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PTextureID, false, false)
- added two new integral types SpriteID and TextureID, both are needed to allow proper serialization as they require something different to be written out than a regular integer.
2016-11-21 12:45:33 +00:00
//==========================================================================
//
// PTextureID Default Constructor
//
//==========================================================================
PTextureID::PTextureID()
: PInt(sizeof(FTextureID), true, false)
{
mDescriptiveName = "TextureID";
assert(sizeof(FTextureID) == alignof(FTextureID));
}
//==========================================================================
//
// PTextureID :: WriteValue
//
//==========================================================================
void PTextureID::WriteValue(FSerializer &ar, const char *key, const void *addr) const
{
FTextureID val = *(FTextureID*)addr;
ar(key, val);
}
//==========================================================================
//
// PTextureID :: ReadValue
//
//==========================================================================
bool PTextureID::ReadValue(FSerializer &ar, const char *key, void *addr) const
{
FTextureID val;
ar(key, val);
*(FTextureID*)addr = val;
return true;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PSound *****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PSound, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PSound Default Constructor
//
//==========================================================================
PSound::PSound()
: PInt(sizeof(FSoundID), true)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Sound";
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
assert(sizeof(FSoundID) == alignof(FSoundID));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PSound :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PSound::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const char *cptr = *(const FSoundID *)addr;
ar.StringPtr(key, cptr);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
//==========================================================================
//
// PSound :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PSound::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const char *cptr;
ar.StringPtr(key, cptr);
if (cptr == nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return false;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
*(FSoundID *)addr = FSoundID(cptr);
return true;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PColor *****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PColor, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PColor Default Constructor
//
//==========================================================================
PColor::PColor()
: PInt(sizeof(PalEntry), true)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Color";
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
assert(sizeof(PalEntry) == alignof(PalEntry));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
/* PStateLabel *****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PStateLabel, false, false)
- fixed: State labels were resolved in the calling function's context instead of the called function one's. This could cause problems with functions that take states as parameters but use them to set them internally instead of passing them through the A_Jump interface back to the caller, like A_Chase or A_LookEx. This required some quite significant refactoring because the entire state resolution logic had been baked into the compiler which turned out to be a major maintenance problem. Fixed this by adding a new builtin type 'statelabel'. This is an opaque identifier representing a state, with the actual data either directly encoded into the number for single label state or an index into a state information table. The state resolution is now the task of the called function as it should always have remained. Note, that this required giving back the 'action' qualifier to most state jumping functions. - refactored most A_Jump checkers to a two stage setup with a pure checker that returns a boolean and a scripted A_Jump wrapper, for some simpler checks the checker function was entirely omitted and calculated inline in the A_Jump function. It is strongly recommended to use the boolean checkers unless using an inline function invocation in a state as they lead to vastly clearer code and offer more flexibility. - let Min() and Max() use the OP_MIN and OP_MAX opcodes. Although these were present, these function were implemented using some grossly inefficient branching tests. - the DECORATE 'state' cast kludge will now actually call ResolveState because a state label is not a state and needs conversion.
2016-11-14 13:12:27 +00:00
//==========================================================================
//
// PStateLabel Default Constructor
//
//==========================================================================
PStateLabel::PStateLabel()
: PInt(sizeof(int), false, false)
{
mDescriptiveName = "StateLabel";
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PPointer ***************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PPointer, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PPointer - Default Constructor
//
//==========================================================================
PPointer::PPointer()
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: PBasicType(sizeof(void *), alignof(void *)), PointedType(nullptr), IsConst(false)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- preparations for null pointer support. - add pointer support to FxAssign.
2016-10-21 17:18:39 +00:00
mDescriptiveName = "NullPointer";
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
SetOps();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PPointer - Parameterized Constructor
//
//==========================================================================
- added readonly pointers. They need to be defined with 'readonly<classtype>'. These are significantly different from declaring a field readonly in that they do not disallow modification of the variable itself but what it points to. For the actor defaults this is necessary to prevent accidental modification. A readonly pointer is actually a different type than a regular pointer. - fixed code generation for dynamic cast. It was missing the jump instruction after the compare.
2016-11-05 12:51:46 +00:00
PPointer::PPointer(PType *pointsat, bool isconst)
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
: PBasicType(sizeof(void *), alignof(void *)), PointedType(pointsat), IsConst(isconst)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- added readonly pointers. They need to be defined with 'readonly<classtype>'. These are significantly different from declaring a field readonly in that they do not disallow modification of the variable itself but what it points to. For the actor defaults this is necessary to prevent accidental modification. A readonly pointer is actually a different type than a regular pointer. - fixed code generation for dynamic cast. It was missing the jump instruction after the compare.
2016-11-05 12:51:46 +00:00
mDescriptiveName.Format("Pointer<%s%s>", pointsat->DescriptiveName(), isconst? "readonly " : "");
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
SetOps();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PPointer :: GetStoreOp
//
//==========================================================================
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
void PPointer::SetOps()
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
loadOp = (PointedType && PointedType->IsKindOf(RUNTIME_CLASS(PClass))) ? OP_LO : OP_LP;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// Non-destroyed thinkers are always guaranteed to be linked into the thinker chain so we don't need the write barrier for them.
storeOp = (loadOp == OP_LO && !static_cast<PClass*>(PointedType)->IsDescendantOf(RUNTIME_CLASS(DThinker))) ? OP_SO : OP_SP;
- changed GetStore/LoadMoveOp and GetRegType to be based on a set of member variables instead of using virtual functions to return such trivial values. This not only creates significantly better code but also allows overriding these settings without having to create a new class just for changing them (as is needed for vectors.)
2016-10-28 08:44:01 +00:00
moveOp = OP_MOVEA;
RegType = REGT_POINTER;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PPointer :: IsMatch
//
//==========================================================================
bool PPointer::IsMatch(intptr_t id1, intptr_t id2) const
{
- added an accessor to the actor defaults. This might have been possible with less work using a function but that would have necessitated some type casts when using it on subclasses. - scriptified A_BarrelDestroy to test the above.
2016-11-05 16:14:16 +00:00
assert(id2 == 0 || id2 == 1);
Normalize line endings
2016-03-01 15:47:10 +00:00
PType *pointat = (PType *)id1;
- added an accessor to the actor defaults. This might have been possible with less work using a function but that would have necessitated some type casts when using it on subclasses. - scriptified A_BarrelDestroy to test the above.
2016-11-05 16:14:16 +00:00
return pointat == PointedType && (!!id2) == IsConst;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PPointer :: GetTypeIDs
//
//==========================================================================
void PPointer::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)PointedType;
id2 = 0;
}
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
//==========================================================================
//
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// PPointer :: SetPointer
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
//
//==========================================================================
void PPointer::SetPointer(void *base, unsigned offset, TArray<size_t> *special) const
{
if (PointedType != nullptr && PointedType->IsKindOf(RUNTIME_CLASS(PClass)))
{
// Add to the list of pointers for this class.
special->Push(offset);
}
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PPointer :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PPointer::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
if (PointedType->IsKindOf(RUNTIME_CLASS(PClass)))
- fixed: PPointer was unable to handle pointers to class types for savegames.
2017-01-02 12:37:09 +00:00
{
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
auto pt = static_cast<PClass*>(PointedType);
if (pt->IsDescendantOf(RUNTIME_CLASS(PClass)))
{
ar(key, *(PClass **)addr);
}
else
{
ar(key, *(DObject **)addr);
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
assert(0 && "Pointer points to a type we don't handle");
I_Error("Attempt to save pointer to unhandled type");
}
}
//==========================================================================
//
// PPointer :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PPointer::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
if (PointedType->IsKindOf(RUNTIME_CLASS(PClass)))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
auto pt = static_cast<PClass*>(PointedType);
bool res = true;
if (pt->IsDescendantOf(RUNTIME_CLASS(PClass)))
{
::Serialize(ar, key, *(PClass **)addr, (PClass**)nullptr);
}
else
{
::Serialize(ar, key, *(DObject **)addr, nullptr, &res);
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return res;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
return false;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// NewPointer
//
// Returns a PPointer to an object of the specified type
//
//==========================================================================
- added readonly pointers. They need to be defined with 'readonly<classtype>'. These are significantly different from declaring a field readonly in that they do not disallow modification of the variable itself but what it points to. For the actor defaults this is necessary to prevent accidental modification. A readonly pointer is actually a different type than a regular pointer. - fixed code generation for dynamic cast. It was missing the jump instruction after the compare.
2016-11-05 12:51:46 +00:00
PPointer *NewPointer(PType *type, bool isconst)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
size_t bucket;
- added readonly pointers. They need to be defined with 'readonly<classtype>'. These are significantly different from declaring a field readonly in that they do not disallow modification of the variable itself but what it points to. For the actor defaults this is necessary to prevent accidental modification. A readonly pointer is actually a different type than a regular pointer. - fixed code generation for dynamic cast. It was missing the jump instruction after the compare.
2016-11-05 12:51:46 +00:00
PType *ptype = TypeTable.FindType(RUNTIME_CLASS(PPointer), (intptr_t)type, isconst ? 1 : 0, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (ptype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- fixed: Readonly pointers never were flagged as such. - fixed: Assignment from a readonly to a read-allowed pointer must be an error. - made GetDefaultByType a builtin so that it can do proper type assignment to the result, which for a function would be problematic in this case, even if automatic type deduction was implemented. Since this returns the class defaults which are not a real object, the result cannot be subjected to a type cast. - error out if a type cast of a readonly pointer is attempted. - fixed: FxBooleanNot could clobber a local variable because it used the source register to manipulate the result.
2016-11-27 14:50:58 +00:00
ptype = new PPointer(type, isconst);
- added readonly pointers. They need to be defined with 'readonly<classtype>'. These are significantly different from declaring a field readonly in that they do not disallow modification of the variable itself but what it points to. For the actor defaults this is necessary to prevent accidental modification. A readonly pointer is actually a different type than a regular pointer. - fixed code generation for dynamic cast. It was missing the jump instruction after the compare.
2016-11-05 12:51:46 +00:00
TypeTable.AddType(ptype, RUNTIME_CLASS(PPointer), (intptr_t)type, isconst ? 1 : 0, bucket);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
return static_cast<PPointer *>(ptype);
}
- made PStatePointer a real pointer pointing to a native struct named 'state', because that's what it really is and what it needs to be if we want to use a state as more than an opaque identifier in the VM
2016-11-21 18:36:14 +00:00
/* PStatePointer **********************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PStatePointer, false, false)
- made PStatePointer a real pointer pointing to a native struct named 'state', because that's what it really is and what it needs to be if we want to use a state as more than an opaque identifier in the VM
2016-11-21 18:36:14 +00:00
//==========================================================================
//
// PStatePointer Default Constructor
//
//==========================================================================
PStatePointer::PStatePointer()
{
mDescriptiveName = "Pointer<State>";
PointedType = NewNativeStruct(NAME_State, nullptr);
IsConst = true;
}
//==========================================================================
//
// PStatePointer :: WriteValue
//
//==========================================================================
void PStatePointer::WriteValue(FSerializer &ar, const char *key, const void *addr) const
{
ar(key, *(FState **)addr);
}
//==========================================================================
//
// PStatePointer :: ReadValue
//
//==========================================================================
bool PStatePointer::ReadValue(FSerializer &ar, const char *key, void *addr) const
{
bool res = false;
::Serialize(ar, key, *(FState **)addr, nullptr, &res);
return res;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PClassPointer **********************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PClassPointer,false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClassPointer - Parameterized Constructor
//
//==========================================================================
PClassPointer::PClassPointer(PClass *restrict)
: PPointer(RUNTIME_CLASS(PClass)), ClassRestriction(restrict)
{
- converted some Heretic stuff to ZScript for testing. - added type casts to the arguments of function calls. - added string constant to state conversion to FxTypeCast.
2016-10-16 20:32:52 +00:00
if (restrict) mDescriptiveName.Format("ClassPointer<%s>", restrict->TypeName.GetChars());
else mDescriptiveName = "ClassPointer";
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// class pointers do not need write barriers because all classes are stored in the global type table and won't get collected.
// This means we can use the cheapoer non-barriered opcodes here.
loadOp = OP_LOS;
storeOp = OP_SP;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- eliminated the native PowerupGiver class. - scriptified the respawn invulnerability code into a virtual OnRespawn function for PlayerPawn so that custom effects can be implemented.
2017-01-17 23:11:04 +00:00
//==========================================================================
//
// PClassPointer - isCompatible
//
//==========================================================================
bool PClassPointer::isCompatible(PType *type)
{
auto other = dyn_cast<PClassPointer>(type);
return (other != nullptr && other->ClassRestriction->IsDescendantOf(ClassRestriction));
}
- fixed: When loading a savegame the player class in the userinfo needs to be updated. - fixed: Class pointers should not be added to the list of garbage collected pointers. - fixed several warnings in the event code.
2017-02-15 10:55:08 +00:00
//==========================================================================
//
// PClassPointer :: SetPointer
//
//==========================================================================
void PClassPointer::SetPointer(void *base, unsigned offset, TArray<size_t> *special) const
{
// Class pointers do not get added to FlatPointers because they are released from the GC.
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClassPointer :: IsMatch
//
//==========================================================================
bool PClassPointer::IsMatch(intptr_t id1, intptr_t id2) const
{
const PType *pointat = (const PType *)id1;
const PClass *classat = (const PClass *)id2;
assert(pointat->IsKindOf(RUNTIME_CLASS(PClass)));
return classat == ClassRestriction;
}
//==========================================================================
//
// PClassPointer :: GetTypeIDs
//
//==========================================================================
void PClassPointer::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
assert(PointedType == RUNTIME_CLASS(PClass));
id1 = (intptr_t)PointedType;
id2 = (intptr_t)ClassRestriction;
}
//==========================================================================
//
// NewClassPointer
//
// Returns a PClassPointer for the restricted type.
//
//==========================================================================
PClassPointer *NewClassPointer(PClass *restrict)
{
size_t bucket;
PClassPointer should use its own type as the metatype in the type table - I don't remember why I thought using PPointer as the metatype for PClassPointer would be preferable, but it means that PPointer's MatchID can potentially be called for PClassPointer entries.
2016-03-03 21:09:19 +00:00
PType *ptype = TypeTable.FindType(RUNTIME_CLASS(PClassPointer), (intptr_t)RUNTIME_CLASS(PClass), (intptr_t)restrict, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (ptype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
ptype = new PClassPointer(restrict);
PClassPointer should use its own type as the metatype in the type table - I don't remember why I thought using PPointer as the metatype for PClassPointer would be preferable, but it means that PPointer's MatchID can potentially be called for PClassPointer entries.
2016-03-03 21:09:19 +00:00
TypeTable.AddType(ptype, RUNTIME_CLASS(PClassPointer), (intptr_t)RUNTIME_CLASS(PClass), (intptr_t)restrict, bucket);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
return static_cast<PClassPointer *>(ptype);
}
/* PEnum ******************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PEnum, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PEnum - Default Constructor
//
//==========================================================================
PEnum::PEnum()
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
: PInt(4, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Enum";
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PEnum - Parameterized Constructor
//
//==========================================================================
- do not declare PNamedType::Outer as DObject. Ultimately we may have to get a fully qualified name out of this, so Outer should be a type that can handle this feature. The new class for this is currently used as base for PType and PSymbol so that PNamedType inherits from it and maybe later a namespace symbol can, too.
2016-10-07 11:59:29 +00:00
PEnum::PEnum(FName name, PTypeBase *outer)
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
: PInt(4, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
EnumName = name;
Outer = outer;
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("Enum<%s>", name.GetChars());
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// NewEnum
//
// Returns a PEnum for the given name and container, making sure not to
// create duplicates.
//
//==========================================================================
- do not declare PNamedType::Outer as DObject. Ultimately we may have to get a fully qualified name out of this, so Outer should be a type that can handle this feature. The new class for this is currently used as base for PType and PSymbol so that PNamedType inherits from it and maybe later a namespace symbol can, too.
2016-10-07 11:59:29 +00:00
PEnum *NewEnum(FName name, PTypeBase *outer)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
size_t bucket;
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
if (outer == nullptr) outer = Namespaces.GlobalNamespace;
Normalize line endings
2016-03-01 15:47:10 +00:00
PType *etype = TypeTable.FindType(RUNTIME_CLASS(PEnum), (intptr_t)outer, (intptr_t)name, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (etype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
etype = new PEnum(name, outer);
TypeTable.AddType(etype, RUNTIME_CLASS(PEnum), (intptr_t)outer, (intptr_t)name, bucket);
}
return static_cast<PEnum *>(etype);
}
/* PArray *****************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PArray, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PArray - Default Constructor
//
//==========================================================================
PArray::PArray()
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: ElementType(nullptr), ElementCount(0)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Array";
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PArray - Parameterized Constructor
//
//==========================================================================
PArray::PArray(PType *etype, unsigned int ecount)
: ElementType(etype), ElementCount(ecount)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("Array<%s>[%d]", etype->DescriptiveName(), ecount);
Normalize line endings
2016-03-01 15:47:10 +00:00
Align = etype->Align;
// Since we are concatenating elements together, the element size should
// also be padded to the nearest alignment.
ElementSize = (etype->Size + (etype->Align - 1)) & ~(etype->Align - 1);
Size = ElementSize * ecount;
}
//==========================================================================
//
// PArray :: IsMatch
//
//==========================================================================
bool PArray::IsMatch(intptr_t id1, intptr_t id2) const
{
const PType *elemtype = (const PType *)id1;
unsigned int count = (unsigned int)(intptr_t)id2;
return elemtype == ElementType && count == ElementCount;
}
//==========================================================================
//
// PArray :: GetTypeIDs
//
//==========================================================================
void PArray::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)ElementType;
id2 = ElementCount;
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PArray :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PArray::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (ar.BeginArray(key))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
const BYTE *addrb = (const BYTE *)addr;
for (unsigned i = 0; i < ElementCount; ++i)
{
ElementType->WriteValue(ar, nullptr, addrb);
addrb += ElementSize;
}
ar.EndArray();
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
//==========================================================================
//
// PArray :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PArray::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (ar.BeginArray(key))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool readsomething = false;
unsigned count = ar.ArraySize();
unsigned loop = MIN(count, ElementCount);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
BYTE *addrb = (BYTE *)addr;
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
for(unsigned i=0;i<loop;i++)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- re-added serialization of treanslations. - removed more dead code.
2016-09-20 17:45:32 +00:00
readsomething |= ElementType->ReadValue(ar, nullptr, addrb);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
addrb += ElementSize;
}
- re-added serialization of treanslations. - removed more dead code.
2016-09-20 17:45:32 +00:00
if (loop < count)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- added multiple message levels for 'developer' CVAR so that the important stuff won't get drowned in pointless notification spam that's of no use to anyone. - made 'developer' CVAR persist across launches and added some menu entries for it. - added checks for 'developer' to ACS's CheckInventory function.
2016-08-28 07:55:04 +00:00
DPrintf(DMSG_WARNING, "Array on disk (%u) is bigger than in memory (%u)\n",
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
count, ElementCount);
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
ar.EndArray();
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
return readsomething;
}
return false;
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PArray :: SetDefaultValue
//
//==========================================================================
void PArray::SetDefaultValue(void *base, unsigned offset, TArray<FTypeAndOffset> *special) const
{
for (unsigned i = 0; i < ElementCount; ++i)
{
ElementType->SetDefaultValue(base, offset + i*ElementSize, special);
}
}
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
//==========================================================================
//
// PArray :: SetDefaultValue
//
//==========================================================================
void PArray::SetPointer(void *base, unsigned offset, TArray<size_t> *special) const
{
for (unsigned i = 0; i < ElementCount; ++i)
{
ElementType->SetPointer(base, offset + i*ElementSize, special);
}
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
// NewArray
Normalize line endings
2016-03-01 15:47:10 +00:00
//
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
// Returns a PArray for the given type and size, making sure not to create
Normalize line endings
2016-03-01 15:47:10 +00:00
// duplicates.
//
//==========================================================================
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
PArray *NewArray(PType *type, unsigned int count)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
size_t bucket;
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
PType *atype = TypeTable.FindType(RUNTIME_CLASS(PArray), (intptr_t)type, count, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (atype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
atype = new PArray(type, count);
TypeTable.AddType(atype, RUNTIME_CLASS(PArray), (intptr_t)type, count, bucket);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- added bookkeeping info to types for allowing data on the stack that requires construction and destruction. - removed unused PVector type.
2016-11-17 10:16:00 +00:00
return (PArray *)atype;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- added a TStaticArray class that allows safe access to resizable static data (like the sectors, linedefs, etc.) for the VM. - used this to replace the line list in Sector because that gets already used and implemented proper bounds checks for this type of array.
2017-01-02 20:40:52 +00:00
/* PArray *****************************************************************/
IMPLEMENT_CLASS(PResizableArray, false, false)
//==========================================================================
//
// PArray - Default Constructor
//
//==========================================================================
PResizableArray::PResizableArray()
{
mDescriptiveName = "ResizableArray";
}
//==========================================================================
//
// PArray - Parameterized Constructor
//
//==========================================================================
PResizableArray::PResizableArray(PType *etype)
: PArray(etype, 0)
{
mDescriptiveName.Format("ResizableArray<%s>", etype->DescriptiveName());
}
//==========================================================================
//
// PArray :: IsMatch
//
//==========================================================================
bool PResizableArray::IsMatch(intptr_t id1, intptr_t id2) const
{
const PType *elemtype = (const PType *)id1;
unsigned int count = (unsigned int)(intptr_t)id2;
return elemtype == ElementType && count == 0;
}
//==========================================================================
//
// PArray :: GetTypeIDs
//
//==========================================================================
void PResizableArray::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)ElementType;
id2 = 0;
}
//==========================================================================
//
// NewResizableArray
//
// Returns a PArray for the given type and size, making sure not to create
// duplicates.
//
//==========================================================================
PResizableArray *NewResizableArray(PType *type)
{
size_t bucket;
PType *atype = TypeTable.FindType(RUNTIME_CLASS(PResizableArray), (intptr_t)type, 0, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (atype == nullptr)
- added a TStaticArray class that allows safe access to resizable static data (like the sectors, linedefs, etc.) for the VM. - used this to replace the line list in Sector because that gets already used and implemented proper bounds checks for this type of array.
2017-01-02 20:40:52 +00:00
{
atype = new PResizableArray(type);
TypeTable.AddType(atype, RUNTIME_CLASS(PResizableArray), (intptr_t)type, 0, bucket);
}
return (PResizableArray *)atype;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PDynArray **************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PDynArray, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PDynArray - Default Constructor
//
//==========================================================================
PDynArray::PDynArray()
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: ElementType(nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "DynArray";
Normalize line endings
2016-03-01 15:47:10 +00:00
Size = sizeof(FArray);
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
Align = alignof(FArray);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PDynArray - Parameterized Constructor
//
//==========================================================================
- major work on dynamic array support. Mostly working, some issues still exist with Move and Copy methods and with assignments in stack variables.
2017-02-06 20:39:21 +00:00
PDynArray::PDynArray(PType *etype,PStruct *backing)
: ElementType(etype), BackingType(backing)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("DynArray<%s>", etype->DescriptiveName());
Normalize line endings
2016-03-01 15:47:10 +00:00
Size = sizeof(FArray);
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
Align = alignof(FArray);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PDynArray :: IsMatch
//
//==========================================================================
bool PDynArray::IsMatch(intptr_t id1, intptr_t id2) const
{
assert(id2 == 0);
const PType *elemtype = (const PType *)id1;
return elemtype == ElementType;
}
//==========================================================================
//
// PDynArray :: GetTypeIDs
//
//==========================================================================
void PDynArray::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)ElementType;
id2 = 0;
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
//==========================================================================
//
// PDynArray :: InitializeValue
//
//==========================================================================
void PDynArray::InitializeValue(void *addr, const void *deff) const
{
const FArray *def = (const FArray*)deff;
FArray *aray = (FArray*)addr;
if (def == nullptr || def->Count == 0)
{
// Empty arrays do not need construction.
*aray = { nullptr, 0, 0 };
}
else if (ElementType->GetRegType() != REGT_STRING)
{
// These are just integral values which can be done without any constructor hackery.
size_t blocksize = ElementType->Size * def->Count;
aray->Array = M_Malloc(blocksize);
memcpy(aray->Array, def->Array, blocksize);
aray->Most = aray->Count = def->Count;
}
else
{
// non-empty string arrays require explicit construction.
new(addr) TArray<FString>(*(TArray<FString>*)def);
}
}
//==========================================================================
//
// PDynArray :: DestroyValue
//
//==========================================================================
void PDynArray::DestroyValue(void *addr) const
{
FArray *aray = (FArray*)addr;
if (aray->Array != nullptr)
{
if (ElementType->GetRegType() != REGT_STRING)
{
M_Free(aray->Array);
}
else
{
// Damn those cursed strings again. :(
((TArray<FString>*)addr)->~TArray<FString>();
}
}
aray->Count = aray->Most = 0;
aray->Array = nullptr;
}
//==========================================================================
//
// PDynArray :: SetDefaultValue
//
//==========================================================================
void PDynArray::SetDefaultValue(void *base, unsigned offset, TArray<FTypeAndOffset> *special) const
{
- everything compiles and mostly works again.
2017-02-12 13:04:48 +00:00
if (base != nullptr) memset((char*)base + offset, 0, sizeof(FArray)); // same as constructing an empty array.
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (special != nullptr)
{
special->Push(std::make_pair(this, offset));
}
}
//==========================================================================
//
// PDynArray :: SetPointer
//
//==========================================================================
void PDynArray::SetPointerArray(void *base, unsigned offset, TArray<size_t> *special) const
{
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
if (ElementType->IsKindOf(RUNTIME_CLASS(PPointer)) && !ElementType->IsKindOf(RUNTIME_CLASS(PClassPointer)) && static_cast<PPointer*>(ElementType)->PointedType->IsKindOf(RUNTIME_CLASS(PClass)))
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
{
// Add to the list of pointer arrays for this class.
special->Push(offset);
}
}
//==========================================================================
//
// PDynArray :: WriteValue
//
//==========================================================================
void PDynArray::WriteValue(FSerializer &ar, const char *key, const void *addr) const
{
FArray *aray = (FArray*)addr;
if (aray->Count > 0)
{
if (ar.BeginArray(key))
{
const BYTE *addrb = (const BYTE *)aray->Array;
for (unsigned i = 0; i < aray->Count; ++i)
{
ElementType->WriteValue(ar, nullptr, addrb);
addrb += ElementType->Size;
}
ar.EndArray();
}
}
}
//==========================================================================
//
// PDynArray :: ReadValue
//
//==========================================================================
bool PDynArray::ReadValue(FSerializer &ar, const char *key, void *addr) const
{
FArray *aray = (FArray*)addr;
DestroyValue(addr); // note that even after calling this we still got a validly constructed empty array.
if (ar.BeginArray(key))
{
bool readsomething = false;
unsigned count = ar.ArraySize();
size_t blocksize = ElementType->Size * count;
aray->Array = M_Malloc(blocksize);
memset(aray->Array, 0, blocksize);
aray->Most = aray->Count = count;
BYTE *addrb = (BYTE *)aray->Array;
for (unsigned i = 0; i<count; i++)
{
// Strings must be constructed first.
if (ElementType->GetRegType() == REGT_STRING) new(addrb) FString;
readsomething |= ElementType->ReadValue(ar, nullptr, addrb);
addrb += ElementType->Size;
}
ar.EndArray();
return readsomething;
}
return false;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// NewDynArray
//
// Creates a new DynArray of the given type, making sure not to create a
// duplicate.
//
//==========================================================================
PDynArray *NewDynArray(PType *type)
{
size_t bucket;
PType *atype = TypeTable.FindType(RUNTIME_CLASS(PDynArray), (intptr_t)type, 0, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (atype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- major work on dynamic array support. Mostly working, some issues still exist with Move and Copy methods and with assignments in stack variables.
2017-02-06 20:39:21 +00:00
FString backingname;
switch (type->GetRegType())
{
case REGT_INT:
backingname.Format("DynArray_I%d", type->Size * 8);
break;
case REGT_FLOAT:
backingname.Format("DynArray_F%d", type->Size * 8);
break;
case REGT_STRING:
backingname = "DynArray_String";
break;
case REGT_POINTER:
backingname = "DynArray_Ptr";
break;
default:
I_Error("Unsupported dynamic array requested");
break;
}
auto backing = NewNativeStruct(backingname, nullptr);
atype = new PDynArray(type, backing);
Normalize line endings
2016-03-01 15:47:10 +00:00
TypeTable.AddType(atype, RUNTIME_CLASS(PDynArray), (intptr_t)type, 0, bucket);
}
return (PDynArray *)atype;
}
/* PMap *******************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PMap, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PMap - Default Constructor
//
//==========================================================================
PMap::PMap()
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: KeyType(nullptr), ValueType(nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Map";
Normalize line endings
2016-03-01 15:47:10 +00:00
Size = sizeof(FMap);
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
Align = alignof(FMap);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PMap - Parameterized Constructor
//
//==========================================================================
PMap::PMap(PType *keytype, PType *valtype)
: KeyType(keytype), ValueType(valtype)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("Map<%s, %s>", keytype->DescriptiveName(), valtype->DescriptiveName());
Normalize line endings
2016-03-01 15:47:10 +00:00
Size = sizeof(FMap);
- script export of player_t. - replaced __alignof with the standard alignof equivalent.
2016-11-17 23:42:04 +00:00
Align = alignof(FMap);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PMap :: IsMatch
//
//==========================================================================
bool PMap::IsMatch(intptr_t id1, intptr_t id2) const
{
const PType *keyty = (const PType *)id1;
const PType *valty = (const PType *)id2;
return keyty == KeyType && valty == ValueType;
}
//==========================================================================
//
// PMap :: GetTypeIDs
//
//==========================================================================
void PMap::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)KeyType;
id2 = (intptr_t)ValueType;
}
//==========================================================================
//
// NewMap
//
// Returns a PMap for the given key and value types, ensuring not to create
// duplicates.
//
//==========================================================================
PMap *NewMap(PType *keytype, PType *valuetype)
{
size_t bucket;
PType *maptype = TypeTable.FindType(RUNTIME_CLASS(PMap), (intptr_t)keytype, (intptr_t)valuetype, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (maptype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
maptype = new PMap(keytype, valuetype);
TypeTable.AddType(maptype, RUNTIME_CLASS(PMap), (intptr_t)keytype, (intptr_t)valuetype, bucket);
}
return (PMap *)maptype;
}
/* PStruct ****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PStruct, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PStruct - Default Constructor
//
//==========================================================================
PStruct::PStruct()
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Struct";
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
Size = 0;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PStruct - Parameterized Constructor
//
//==========================================================================
- do not declare PNamedType::Outer as DObject. Ultimately we may have to get a fully qualified name out of this, so Outer should be a type that can handle this feature. The new class for this is currently used as base for PType and PSymbol so that PNamedType inherits from it and maybe later a namespace symbol can, too.
2016-10-07 11:59:29 +00:00
PStruct::PStruct(FName name, PTypeBase *outer)
Normalize line endings
2016-03-01 15:47:10 +00:00
: PNamedType(name, outer)
{
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName.Format("Struct<%s>", name.GetChars());
- fixed setup for PString which was missing the load/store/move instructions which caused me to overlook its register type. - properly set up the vector types. - fixed: a struct must start with a size of 0, not 1. This caused the layout of the vectors to be broken.
2016-10-28 12:43:29 +00:00
Size = 0;
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
HasNativeFields = false;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PStruct :: SetDefaultValue
//
//==========================================================================
void PStruct::SetDefaultValue(void *base, unsigned offset, TArray<FTypeAndOffset> *special) const
{
for (const PField *field : Fields)
{
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
if (!(field->Flags & VARF_Transient))
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
field->Type->SetDefaultValue(base, unsigned(offset + field->Offset), special);
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
}
}
}
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
//==========================================================================
//
// PStruct :: SetPointer
//
//==========================================================================
void PStruct::SetPointer(void *base, unsigned offset, TArray<size_t> *special) const
{
for (const PField *field : Fields)
{
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
if (!(field->Flags & VARF_Transient))
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
{
field->Type->SetPointer(base, unsigned(offset + field->Offset), special);
}
}
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PStruct :: WriteValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PStruct::WriteValue(FSerializer &ar, const char *key,const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (ar.BeginObject(key))
{
WriteFields(ar, addr, Fields);
ar.EndObject();
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
//==========================================================================
//
// PStruct :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PStruct::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (ar.BeginObject(key))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool ret = ReadFields(ar, addr);
ar.EndObject();
return ret;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return false;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
//==========================================================================
//
// PStruct :: WriteFields STATIC
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PStruct::WriteFields(FSerializer &ar, const void *addr, const TArray<PField *> &fields)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
for (unsigned i = 0; i < fields.Size(); ++i)
{
const PField *field = fields[i];
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
// Skip fields without or with native serialization
if (!(field->Flags & VARF_Transient))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
field->Type->WriteValue(ar, field->SymbolName.GetChars(), (const BYTE *)addr + field->Offset);
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
}
}
//==========================================================================
//
// PStruct :: ReadFields
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PStruct::ReadFields(FSerializer &ar, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
bool readsomething = false;
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool foundsomething = false;
const char *label;
while ((label = ar.GetKey()))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
foundsomething = true;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
const PSymbol *sym = Symbols.FindSymbol(FName(label, true), true);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (sym == nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- added multiple message levels for 'developer' CVAR so that the important stuff won't get drowned in pointless notification spam that's of no use to anyone. - made 'developer' CVAR persist across launches and added some menu entries for it. - added checks for 'developer' to ACS's CheckInventory function.
2016-08-28 07:55:04 +00:00
DPrintf(DMSG_ERROR, "Cannot find field %s in %s\n",
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
label, TypeName.GetChars());
}
else if (!sym->IsKindOf(RUNTIME_CLASS(PField)))
{
- added multiple message levels for 'developer' CVAR so that the important stuff won't get drowned in pointless notification spam that's of no use to anyone. - made 'developer' CVAR persist across launches and added some menu entries for it. - added checks for 'developer' to ACS's CheckInventory function.
2016-08-28 07:55:04 +00:00
DPrintf(DMSG_ERROR, "Symbol %s in %s is not a field\n",
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
label, TypeName.GetChars());
}
else
{
- removed the sequential processing of JSON objects because the benefit is too small. After testing with a savegame on ZDCMP2 which is probably the largest map in existence, timing both methods resulted in a speed difference of less than 40 ms (70 vs 110 ms for reading all sectory, linedefs, sidedefs and objects). This compares to an overall restoration time, including reloading the level, precaching all textures and setting everything up, of approx. 1.2 s, meaning an increase of 3% of the entire reloading time. That's simply not worth all the negative side effects that may happen with a method that highly depends on proper code construction. On the other hand, using random access means that a savegame version change is only needed now when the semantics of a field change, but not if some get added or deleted. - do not I_Error out in the serializer unless caused by a programming error. It is better to let the serializer finish, collect all the errors and I_Error out when the game is known to be in a stable enough state to allow unwinding.
2016-09-23 12:04:05 +00:00
readsomething |= static_cast<const PField *>(sym)->Type->ReadValue(ar, nullptr,
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
(BYTE *)addr + static_cast<const PField *>(sym)->Offset);
}
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return readsomething || !foundsomething;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PStruct :: AddField
//
// Appends a new field to the end of a struct. Returns either the new field
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// or nullptr if a symbol by that name already exists.
Normalize line endings
2016-03-01 15:47:10 +00:00
//
//==========================================================================
PField *PStruct::AddField(FName name, PType *type, DWORD flags)
{
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
PField *field = new PField(name, type, flags);
Normalize line endings
2016-03-01 15:47:10 +00:00
// The new field is added to the end of this struct, alignment permitting.
field->Offset = (Size + (type->Align - 1)) & ~(type->Align - 1);
// Enlarge this struct to enclose the new field.
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
Size = unsigned(field->Offset + type->Size);
Normalize line endings
2016-03-01 15:47:10 +00:00
// This struct's alignment is the same as the largest alignment of any of
// its fields.
Align = MAX(Align, type->Align);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (Symbols.AddSymbol(field) == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{ // name is already in use
- remove all symbols that get linked into the symbol table from the garbage collector. Symbols are very easy to manage once they are in a symbol table and there's lots of them so this reduces the amount of work the GC needs to do quite considerably. After cleaning out compile-time-only symbols there will still be more than 2000 left, one for each function and one for each member variable of a class or struct. This means more than 2000 object that won't need to tracked constantly by the garbage collector. Note that loose fields which do occur during code generation will be GC'd just as before.
2017-02-08 13:34:39 +00:00
field->Destroy();
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
Fields.Push(field);
return field;
}
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
//==========================================================================
//
// PStruct :: AddField
//
// Appends a new native field to the struct. Returns either the new field
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// or nullptr if a symbol by that name already exists.
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
//
//==========================================================================
PField *PStruct::AddNativeField(FName name, PType *type, size_t address, DWORD flags, int bitvalue)
{
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
PField *field = new PField(name, type, flags|VARF_Native|VARF_Transient, address, bitvalue);
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
if (Symbols.AddSymbol(field) == nullptr)
{ // name is already in use
- do not delete duplicate PFields - they are better be destroyed and left to the GC. - removed two duplicate entries to the flag table.
2016-11-04 14:21:45 +00:00
field->Destroy();
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
return nullptr;
}
Fields.Push(field);
HasNativeFields = true;
return field;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// NewStruct
// Returns a PStruct for the given name and container, making sure not to
// create duplicates.
//
//==========================================================================
- do not declare PNamedType::Outer as DObject. Ultimately we may have to get a fully qualified name out of this, so Outer should be a type that can handle this feature. The new class for this is currently used as base for PType and PSymbol so that PNamedType inherits from it and maybe later a namespace symbol can, too.
2016-10-07 11:59:29 +00:00
PStruct *NewStruct(FName name, PTypeBase *outer)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
size_t bucket;
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
if (outer == nullptr) outer = Namespaces.GlobalNamespace;
Normalize line endings
2016-03-01 15:47:10 +00:00
PType *stype = TypeTable.FindType(RUNTIME_CLASS(PStruct), (intptr_t)outer, (intptr_t)name, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (stype == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
stype = new PStruct(name, outer);
TypeTable.AddType(stype, RUNTIME_CLASS(PStruct), (intptr_t)outer, (intptr_t)name, bucket);
}
return static_cast<PStruct *>(stype);
}
- allow functions in structs. This is needed for several internal data types like players, sectors, lines, etc. - added a new type 'NativeStruct'. This will be used for types that cannot be instantiated, and is also needed to cleanly handle many internal types that only can exist as reference.
2016-11-19 11:12:29 +00:00
/* PNativeStruct ****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PNativeStruct, false, false)
- allow functions in structs. This is needed for several internal data types like players, sectors, lines, etc. - added a new type 'NativeStruct'. This will be used for types that cannot be instantiated, and is also needed to cleanly handle many internal types that only can exist as reference.
2016-11-19 11:12:29 +00:00
//==========================================================================
//
// PNativeStruct - Parameterized Constructor
//
//==========================================================================
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
PNativeStruct::PNativeStruct(FName name, PTypeBase *outer)
: PStruct(name, outer)
- allow functions in structs. This is needed for several internal data types like players, sectors, lines, etc. - added a new type 'NativeStruct'. This will be used for types that cannot be instantiated, and is also needed to cleanly handle many internal types that only can exist as reference.
2016-11-19 11:12:29 +00:00
{
mDescriptiveName.Format("NativeStruct<%s>", name.GetChars());
Size = 0;
HasNativeFields = true;
}
- made all internal struct types NativeStructs so that all variables of their types are made references.
2016-11-21 18:20:27 +00:00
//==========================================================================
//
// NewNativeStruct
// Returns a PNativeStruct for the given name and container, making sure not to
// create duplicates.
//
//==========================================================================
PNativeStruct *NewNativeStruct(FName name, PTypeBase *outer)
{
size_t bucket;
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
if (outer == nullptr) outer = Namespaces.GlobalNamespace;
- made all internal struct types NativeStructs so that all variables of their types are made references.
2016-11-21 18:20:27 +00:00
PType *stype = TypeTable.FindType(RUNTIME_CLASS(PNativeStruct), (intptr_t)outer, (intptr_t)name, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (stype == nullptr)
- made all internal struct types NativeStructs so that all variables of their types are made references.
2016-11-21 18:20:27 +00:00
{
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
stype = new PNativeStruct(name, outer);
- made all internal struct types NativeStructs so that all variables of their types are made references.
2016-11-21 18:20:27 +00:00
TypeTable.AddType(stype, RUNTIME_CLASS(PNativeStruct), (intptr_t)outer, (intptr_t)name, bucket);
}
return static_cast<PNativeStruct *>(stype);
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PField *****************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PField, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PField - Default Constructor
//
//==========================================================================
PField::PField()
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
: PSymbol(NAME_None), Offset(0), Type(nullptr), Flags(0)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
}
- moved declaration of native fields into the respective class definitions. This bypasses a declaration in the script in favor of a simpler implementation. In order to work it is always necessary to have an offset table to map the variables to, but doing it fully on the native side only requires adding the type to the declaration.
2016-11-03 23:19:36 +00:00
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
PField::PField(FName name, PType *type, DWORD flags, size_t offset, int bitvalue)
- scriptified A_SpawnFly. - added support for global variables to the code generator - not the compiler, though. For the handful of entries this is needed for it may just as well be done manually. So far FLevelLocals level is the only one being exported. - fixed: The VM disassembler truncated 64 bit pointers to 15 digits because the output buffer was too small. - resolve entire FxSequences instead of aborting on the first failed entry. This allows to output all errors at once.
2016-11-03 12:38:40 +00:00
: PSymbol(name), Offset(offset), Type(type), Flags(flags)
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
{
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
if (bitvalue != 0)
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
{
- moved declaration of native fields into the respective class definitions. This bypasses a declaration in the script in favor of a simpler implementation. In order to work it is always necessary to have an offset table to map the variables to, but doing it fully on the native side only requires adding the type to the declaration.
2016-11-03 23:19:36 +00:00
BitValue = 0;
unsigned val = bitvalue;
while ((val >>= 1)) BitValue++;
if (type->IsA(RUNTIME_CLASS(PInt)) && unsigned(BitValue) < 8u * type->Size)
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
{
// map to the single bytes in the actual variable. The internal bit instructions operate on 8 bit values.
#ifndef __BIG_ENDIAN__
Offset += BitValue / 8;
#else
Offset += type->Size - 1 - BitValue / 8;
#endif
BitValue &= 7;
Type = TypeBool;
}
else
{
// Just abort. Bit fields should only be defined internally.
- added proper definitions for all exported native fields. - synthesize native fields for all declared flags, not just for AActor.
2016-11-22 22:42:32 +00:00
I_Error("Trying to create an invalid bit field element: %s", name.GetChars());
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
}
}
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
else BitValue = -1;
- implement flag variables with the VM's sbit and lbit instructions. - synthesize PField entries from the flag list for AActor. This intentionally excludes the bounce flags for now. - allow deprecated flags that do not call the deprecated flag handler. - disallow constructs like (a = b) = c by not allowing an address request on an assignment operation. - restrict modify/assign on boolean variables to the bit operators. Everything else needs to promote the result to an integer to make sense so it should be disallowed.
2016-10-24 15:18:20 +00:00
}
- added property definitions to the ZScript parser. This will allow defining custom properties for the default block in custom base classes. See 'Health' for an example. - support transient object member variables for information that does not need to be put in a savegame. - fixed: special initialization of objects needs to pass the proper defaults along, otherwise the parent classes will use their own, inappropriate one.
2017-01-15 15:55:30 +00:00
/* PProperty *****************************************************************/
IMPLEMENT_CLASS(PProperty, false, false)
//==========================================================================
//
// PField - Default Constructor
//
//==========================================================================
PProperty::PProperty()
: PSymbol(NAME_None)
{
}
PProperty::PProperty(FName name, TArray<PField *> &fields)
: PSymbol(name)
{
Variables = std::move(fields);
}
Normalize line endings
2016-03-01 15:47:10 +00:00
/* PPrototype *************************************************************/
- disabled the scripted virtual function module after finding out that it only works if each single class that may serve as a parent for scripting is explicitly declared. Needless to say, this is simply too volatile and would require constant active maintenance, not to mention a huge amount of work up front to get going. It also hid a nasty problem with the Destroy method. Due to the way the garbage collector works, Destroy cannot be exposed to scripts as-is. It may be called from scripts but it may not be overridden from scripts because the garbage collector can call this function after all data needed for calling a scripted override has already been destroyed because if that data is also being collected there is no guarantee that proper order of destruction is observed. So for now Destroy is just a normal native method to scripted classes
2016-11-24 20:36:02 +00:00
IMPLEMENT_CLASS(PPrototype, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PPrototype - Default Constructor
//
//==========================================================================
PPrototype::PPrototype()
{
}
//==========================================================================
//
// PPrototype - Parameterized Constructor
//
//==========================================================================
PPrototype::PPrototype(const TArray<PType *> &rettypes, const TArray<PType *> &argtypes)
: ArgumentTypes(argtypes), ReturnTypes(rettypes)
{
}
//==========================================================================
//
// PPrototype :: IsMatch
//
//==========================================================================
bool PPrototype::IsMatch(intptr_t id1, intptr_t id2) const
{
const TArray<PType *> *args = (const TArray<PType *> *)id1;
const TArray<PType *> *rets = (const TArray<PType *> *)id2;
return *args == ArgumentTypes && *rets == ReturnTypes;
}
//==========================================================================
//
// PPrototype :: GetTypeIDs
//
//==========================================================================
void PPrototype::GetTypeIDs(intptr_t &id1, intptr_t &id2) const
{
id1 = (intptr_t)&ArgumentTypes;
id2 = (intptr_t)&ReturnTypes;
}
//==========================================================================
//
// PPrototype :: PropagateMark
//
//==========================================================================
size_t PPrototype::PropagateMark()
{
GC::MarkArray(ArgumentTypes);
GC::MarkArray(ReturnTypes);
return (ArgumentTypes.Size() + ReturnTypes.Size()) * sizeof(void*) +
Super::PropagateMark();
}
//==========================================================================
//
// NewPrototype
//
// Returns a PPrototype for the given return and argument types, making sure
// not to create duplicates.
//
//==========================================================================
PPrototype *NewPrototype(const TArray<PType *> &rettypes, const TArray<PType *> &argtypes)
{
size_t bucket;
PType *proto = TypeTable.FindType(RUNTIME_CLASS(PPrototype), (intptr_t)&argtypes, (intptr_t)&rettypes, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (proto == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
proto = new PPrototype(rettypes, argtypes);
TypeTable.AddType(proto, RUNTIME_CLASS(PPrototype), (intptr_t)&argtypes, (intptr_t)&rettypes, bucket);
}
return static_cast<PPrototype *>(proto);
}
/* PClass *****************************************************************/
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
IMPLEMENT_CLASS(PClass, false, false)
Normalize line endings
2016-03-01 15:47:10 +00:00
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PClass :: WriteValue
//
// Similar to PStruct's version, except it also needs to traverse parent
// classes.
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
static void RecurseWriteFields(const PClass *type, FSerializer &ar, const void *addr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (type != nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
RecurseWriteFields(type->ParentClass, ar, addr);
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
// Don't write this part if it has no non-transient variables
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
for (unsigned i = 0; i < type->Fields.Size(); ++i)
{
- allow defining native fields through scripts. Internally this only requires exporting the address, but not the entire field. - added new VARF_Transient flag so that the decision whether to serialize a field does not depend solely on its native status. It may actually make a lot of sense to use the auto-serializer for native fields, too, as this would eliminate a lot of maintenance code. - defined (u)int8/16 as aliases to the byte and short types (Can't we not just get rid of this naming convention already...?) - exporting the fields of Actor revealed a few name clashes between them and some global types, so Actor.Sector was renamed to CurSector and Actor.Inventory was renamed to Actor.Inv.
2016-11-22 18:20:31 +00:00
if (!(type->Fields[i]->Flags & VARF_Transient))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
// Tag this section with the class it came from in case
// a more-derived class has variables that shadow a less-
// derived class. Whether or not that is a language feature
// that will actually be allowed remains to be seen.
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
FString key;
key.Format("class:%s", type->TypeName.GetChars());
if (ar.BeginObject(key.GetChars()))
{
PStruct::WriteFields(ar, addr, type->Fields);
ar.EndObject();
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
break;
}
}
}
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
void PClass::WriteValue(FSerializer &ar, const char *key,const void *addr) const
{
if (ar.BeginObject(key))
{
RecurseWriteFields(this, ar, addr);
ar.EndObject();
}
}
// Same as WriteValue, but does not create a new object in the serializer
// This is so that user variables do not contain unnecessary subblocks.
void PClass::WriteAllFields(FSerializer &ar, const void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
RecurseWriteFields(this, ar, addr);
}
//==========================================================================
//
// PClass :: ReadValue
//
//==========================================================================
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool PClass::ReadValue(FSerializer &ar, const char *key, void *addr) const
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (ar.BeginObject(key))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
bool ret = ReadAllFields(ar, addr);
ar.EndObject();
return ret;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return true;
}
bool PClass::ReadAllFields(FSerializer &ar, void *addr) const
{
bool readsomething = false;
bool foundsomething = false;
const char *key;
- removed the sequential processing of JSON objects because the benefit is too small. After testing with a savegame on ZDCMP2 which is probably the largest map in existence, timing both methods resulted in a speed difference of less than 40 ms (70 vs 110 ms for reading all sectory, linedefs, sidedefs and objects). This compares to an overall restoration time, including reloading the level, precaching all textures and setting everything up, of approx. 1.2 s, meaning an increase of 3% of the entire reloading time. That's simply not worth all the negative side effects that may happen with a method that highly depends on proper code construction. On the other hand, using random access means that a savegame version change is only needed now when the semantics of a field change, but not if some get added or deleted. - do not I_Error out in the serializer unless caused by a programming error. It is better to let the serializer finish, collect all the errors and I_Error out when the game is known to be in a stable enough state to allow unwinding.
2016-09-23 12:04:05 +00:00
key = ar.GetKey();
if (strcmp(key, "classtype"))
{
// this does not represent a DObject
Printf(TEXTCOLOR_RED "trying to read user variables but got a non-object (first key is '%s')", key);
ar.mErrors++;
return false;
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
while ((key = ar.GetKey()))
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (strncmp(key, "class:", 6))
{
- removed the sequential processing of JSON objects because the benefit is too small. After testing with a savegame on ZDCMP2 which is probably the largest map in existence, timing both methods resulted in a speed difference of less than 40 ms (70 vs 110 ms for reading all sectory, linedefs, sidedefs and objects). This compares to an overall restoration time, including reloading the level, precaching all textures and setting everything up, of approx. 1.2 s, meaning an increase of 3% of the entire reloading time. That's simply not worth all the negative side effects that may happen with a method that highly depends on proper code construction. On the other hand, using random access means that a savegame version change is only needed now when the semantics of a field change, but not if some get added or deleted. - do not I_Error out in the serializer unless caused by a programming error. It is better to let the serializer finish, collect all the errors and I_Error out when the game is known to be in a stable enough state to allow unwinding.
2016-09-23 12:04:05 +00:00
// We have read all user variable blocks.
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
break;
}
foundsomething = true;
PClass *type = PClass::FindClass(key + 6);
if (type != nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
// Only read it if the type is related to this one.
const PClass *parent;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
for (parent = this; parent != nullptr; parent = parent->ParentClass)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
if (parent == type)
{
break;
}
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
if (parent != nullptr)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
- removed the sequential processing of JSON objects because the benefit is too small. After testing with a savegame on ZDCMP2 which is probably the largest map in existence, timing both methods resulted in a speed difference of less than 40 ms (70 vs 110 ms for reading all sectory, linedefs, sidedefs and objects). This compares to an overall restoration time, including reloading the level, precaching all textures and setting everything up, of approx. 1.2 s, meaning an increase of 3% of the entire reloading time. That's simply not worth all the negative side effects that may happen with a method that highly depends on proper code construction. On the other hand, using random access means that a savegame version change is only needed now when the semantics of a field change, but not if some get added or deleted. - do not I_Error out in the serializer unless caused by a programming error. It is better to let the serializer finish, collect all the errors and I_Error out when the game is known to be in a stable enough state to allow unwinding.
2016-09-23 12:04:05 +00:00
if (ar.BeginObject(nullptr))
{
readsomething |= type->ReadFields(ar, addr);
ar.EndObject();
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
else
{
- added multiple message levels for 'developer' CVAR so that the important stuff won't get drowned in pointless notification spam that's of no use to anyone. - made 'developer' CVAR persist across launches and added some menu entries for it. - added checks for 'developer' to ACS's CheckInventory function.
2016-08-28 07:55:04 +00:00
DPrintf(DMSG_ERROR, "Unknown superclass %s of class %s\n",
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
type->TypeName.GetChars(), TypeName.GetChars());
}
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
else
{
DPrintf(DMSG_ERROR, "Unknown superclass %s of class %s\n",
key+6, TypeName.GetChars());
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
- converted the user variable serializer.
2016-09-20 16:27:47 +00:00
return readsomething || !foundsomething;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// cregcmp
//
// Sorter to keep built-in types in a deterministic order. (Needed?)
//
//==========================================================================
- removed STACK_ARGS. The only reason this even existed was that ZDoom's original VC projects used __fastcall. The CMake generated project do not, they stick to __cdecl. Since no performance gain can be seen by using __fastcall the best course of action is to just remove all traces of it from the source and forget that it ever existed.
2016-04-11 08:46:30 +00:00
static int cregcmp (const void *a, const void *b) NO_SANITIZE
Normalize line endings
2016-03-01 15:47:10 +00:00
{
const PClass *class1 = *(const PClass **)a;
const PClass *class2 = *(const PClass **)b;
return strcmp(class1->TypeName, class2->TypeName);
}
//==========================================================================
//
// PClass :: StaticInit STATIC
//
// Creates class metadata for all built-in types.
//
//==========================================================================
void PClass::StaticInit ()
{
atterm (StaticShutdown);
StaticBootstrap();
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
Namespaces.GlobalNamespace = Namespaces.NewNamespace(0);
Normalize line endings
2016-03-01 15:47:10 +00:00
FAutoSegIterator probe(CRegHead, CRegTail);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
while (*++probe != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
((ClassReg *)*probe)->RegisterClass ();
}
// Keep built-in classes in consistant order. I did this before, though
// I'm not sure if this is really necessary to maintain any sort of sync.
qsort(&AllClasses[0], AllClasses.Size(), sizeof(AllClasses[0]), cregcmp);
- implemented complete resolving of constants - both global and class-local. This uses a different algorithm as the old implementation - instead of recursively resolving unknown symbols it will first collect all constants from all scopes and then process them in one operation, doing multiple passes over the list until no more constants can be resolved anymore.
2016-10-08 20:16:10 +00:00
// Set all symbol table relations here so that they are valid right from the start.
for (auto c : AllClasses)
{
if (c->ParentClass != nullptr)
{
c->Symbols.SetParentTable(&c->ParentClass->Symbols);
}
}
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PClass :: StaticShutdown STATIC
//
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// Frees all static class data.
Normalize line endings
2016-03-01 15:47:10 +00:00
//
//==========================================================================
void PClass::StaticShutdown ()
{
- fixed: the restart CCMD must clear out all statically stored pointers to VM functions because they do not survive the shutdown that is needed before loading new data.
2017-01-31 12:41:23 +00:00
// delete all variables containing pointers to script functions.
for (auto p : FunctionPtrList)
{
*p = nullptr;
}
FunctionPtrList.Clear();
- took VMFunction out of the DObject hierarchy. As it stood, just compiling the internal ZScript code created more than 9000 DObjects, none of which really need to be subjected to garbage collection, aside from allowing lazy deallocation. This puts an incredible drag on the garbage collector which often needs several minutes to finish processing before actual deletion can start. The VM functions with roughly 1800 of these objects were by far the easiest to refactor so they are now. They also use a memory arena now which significantly reduces their memory footprint.
2017-02-08 10:13:41 +00:00
VMFunction::DeleteAll();
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// Make a full garbage collection here so that all destroyed but uncollected higher level objects
// that still exist are properly taken down before the low level data is deleted.
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
GC::FullGC();
// From this point onward no scripts may be called anymore because the data needed by the VM is getting deleted now.
// This flags DObject::Destroy not to call any scripted OnDestroy methods anymore.
- scriptified the sector actions.
2017-01-13 12:51:47 +00:00
bVMOperational = false;
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
- to avoid problems with the final garbage collection, the players' PendingWeapon needs to be cleared manually. This is because it can point to a non-standard value which the garbage collector cannot deal with during engine shutdown.
2017-02-09 00:34:07 +00:00
// PendingWeapon must be cleared manually because it is not subjected to the GC if it contains WP_NOCHANGE, which is just RUNTIME_CLASS(AWWeapon).
// But that will get cleared here, confusing the GC if the value is left in.
for (auto &p : players)
{
p.PendingWeapon = nullptr;
}
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
// Unless something went wrong, anything left here should be class and type objects only, which do not own any scripts.
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
bShutdown = true;
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
TypeTable.Clear();
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
Namespaces.ReleaseSymbols();
- Turned DropItem into a plain struct again like it was before the scripting branch got merged. Making this an object had little to no advantage, except being able to remove the deleter code. Now, with some of the class data already being allocated in a memory arena so that freeing it is easier, this can also be used for the drop item lists which makes it unnecessary to subject them to the GC. This also merges the memory arenas for VM functions and flat pointers because both get deleted at the same time so they can share the same one.
2017-02-08 19:37:22 +00:00
ClassDataAllocator.FreeAllBlocks();
Normalize line endings
2016-03-01 15:47:10 +00:00
AllClasses.Clear();
PClassActor::AllActorClasses.Clear();
FAutoSegIterator probe(CRegHead, CRegTail);
- fixed: PClass::StaticShutdown did not remove the class pointers for VM base classes from their class registration entries.
2016-11-12 09:17:57 +00:00
while (*++probe != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- fixed: PClass::StaticShutdown did not remove the class pointers for VM base classes from their class registration entries.
2016-11-12 09:17:57 +00:00
auto cr = ((ClassReg *)*probe);
cr->MyClass = nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- split DObject::Destroy into the main method, a native OnDestroy and a scripted OnDestroy method and made the main method non-virtual This was done to ensure it can be properly overridden in scripts without causing problems when called during engine shutdown for the type and symbol objects the VM needs to work and to have the scripted version always run first. Since the scripted OnDestroy method never calls the native version - the native one is run after the scripted one - this can be simply skipped over during shutdown.
2017-01-12 21:49:18 +00:00
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PClass :: StaticBootstrap STATIC
//
//==========================================================================
void PClass::StaticBootstrap()
{
- removed PClassType and PClassClass. All non-actors now use PClass exclusively as their type descriptor. Getting rid of these two classes already removes a lot of obtuse code from the type system, but there's still three more classes to go before a major cleanup can be undertaken.
2017-02-07 19:45:56 +00:00
PClass *cls = new PClass;
Normalize line endings
2016-03-01 15:47:10 +00:00
PClass::RegistrationInfo.SetupClass(cls);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
// The PClassClass constructor initialized these to nullptr, because the
Normalize line endings
2016-03-01 15:47:10 +00:00
// PClass metadata had not been created yet. Now it has, so we know what
// they should be and can insert them into the type table successfully.
cls->InsertIntoHash();
// Create parent objects before we go so that these definitions are complete.
cls->ParentClass = PClass::RegistrationInfo.ParentType->RegisterClass();
}
//==========================================================================
//
// PClass Constructor
//
//==========================================================================
PClass::PClass()
{
Size = sizeof(DObject);
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
ParentClass = nullptr;
Pointers = nullptr;
FlatPointers = nullptr;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
ArrayPointers = nullptr;
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
HashNext = nullptr;
Defaults = nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
bRuntimeClass = false;
- scriptified parts of a_bossbrain.cpp. Some things cannot be done yet, the script code is there but commented out. - exported thinker iterator and drop item chain to scripting. Unlike its native counterpart the script-side iterator is wrapped into a DObject to allow proper handling for memory management. - fixed: The VMFunctionBuilder only distinguished between member and action functions but failed on static ones. - fixed: FxAssign did not add all needed type casts. Except for purely numeric types it will now wrap the expression in an FxTypeCast. Numeric handling remains unchanged for both performance reasons and not altering semantics for DECORATE. - exported all internal flags as variables to scripting. They still cannot be used in an actor definition. - make ATAG_STATE the same as ATAG_GENERIC. Since state pointers exist as actual variables they can take both values which on occasion can trigger some asserts. - gave PClass a bExported flag, so that scripts cannot see purely internal classes. Especially the types like PInt can cause problems. Todo: we need readonly references to safely expose the actor defaults. Right now some badly behaving code could overwrite them.
2016-10-31 16:02:47 +00:00
bExported = false;
- try to preserve a bit more information about incorrect use of user variables to print more meaningful error messages. This is not complete yet and will need integration with the previous commit.
2016-11-13 11:02:41 +00:00
bDecorateClass = false;
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
ConstructNative = nullptr;
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
Meta = nullptr;
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
mDescriptiveName = "Class";
Normalize line endings
2016-03-01 15:47:10 +00:00
PClass::AllClasses.Push(this);
}
//==========================================================================
//
// PClass Destructor
//
//==========================================================================
PClass::~PClass()
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (Defaults != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
M_Free(Defaults);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
Defaults = nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
if (Meta != nullptr)
{
M_Free(Meta);
Meta = nullptr;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// ClassReg :: RegisterClass
//
// Create metadata describing the built-in class this struct is intended
// for.
//
//==========================================================================
PClass *ClassReg::RegisterClass()
{
static ClassReg *const metaclasses[] =
{
&PClass::RegistrationInfo,
&PClassActor::RegistrationInfo,
};
// Skip classes that have already been registered
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (MyClass != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
return MyClass;
}
// Add type to list
PClass *cls;
if (MetaClassNum >= countof(metaclasses))
{
assert(0 && "Class registry has an invalid meta class identifier");
}
Provide a way to expose a native class to zscript for virtual functions overriding
2016-11-06 10:28:01 +00:00
if (metaclasses[MetaClassNum]->MyClass == nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{ // Make sure the meta class is already registered before registering this one
metaclasses[MetaClassNum]->RegisterClass();
}
cls = static_cast<PClass *>(metaclasses[MetaClassNum]->MyClass->CreateNew());
SetupClass(cls);
cls->InsertIntoHash();
Provide a way to expose a native class to zscript for virtual functions overriding
2016-11-06 10:28:01 +00:00
if (ParentType != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
cls->ParentClass = ParentType->RegisterClass();
}
return cls;
}
//==========================================================================
//
// ClassReg :: SetupClass
//
// Copies the class-defining parameters from a ClassReg to the Class object
// created for it.
//
//==========================================================================
void ClassReg::SetupClass(PClass *cls)
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(MyClass == nullptr);
Normalize line endings
2016-03-01 15:47:10 +00:00
MyClass = cls;
cls->TypeName = FName(Name+1);
cls->Size = SizeOf;
cls->Pointers = Pointers;
cls->ConstructNative = ConstructNative;
- changed order of script files to match the old DECORATE list for easy comparison of the disassembly. - added a descriptive name to all types for error messages. - added a generic type cast node to the code generator. - added a few more cast operations to the 'cast' VM instruction. - extended FxClassTypeCast to handle all possible input that can be cast to a class pointer, not just names.
2016-10-16 17:42:22 +00:00
cls->mDescriptiveName.Format("Class<%s>", cls->TypeName.GetChars());
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PClass :: InsertIntoHash
//
// Add class to the type table.
//
//==========================================================================
void PClass::InsertIntoHash ()
{
size_t bucket;
PType *found;
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
found = TypeTable.FindType(RUNTIME_CLASS(PClass), 0, TypeName, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (found != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{ // This type has already been inserted
- changed handling of duplicate classes in DECORATE. Instead of replacing the original, the second class will get renamed now, using the originating file as an identifier. In the vast majority of cases this should do exactly what is needed: Create an unconflicting second class that can coexist with the original. Unless the class is used by name this should eliminate all problems with this, but so far I haven't seen anything that used them by name. This is choosing the lesser of two evils. While some mod out there may get broken, the old setup meant that the first class of a given name could not be written out to a savegame because it was not retrievable when loading it back.
2016-10-07 13:28:25 +00:00
I_Error("Tried to register class '%s' more than once.\n", TypeName.GetChars());
Normalize line endings
2016-03-01 15:47:10 +00:00
}
else
{
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
TypeTable.AddType(this, RUNTIME_CLASS(PClass), 0, TypeName, bucket);
Normalize line endings
2016-03-01 15:47:10 +00:00
}
}
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
//==========================================================================
//
// PClass :: FindParentClass
//
// Finds a parent class that matches the given name, including itself.
//
//==========================================================================
const PClass *PClass::FindParentClass(FName name) const
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
for (const PClass *type = this; type != nullptr; type = type->ParentClass)
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
{
if (type->TypeName == name)
{
return type;
}
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return nullptr;
Added methods for PTypes to serialize their values - Values are tagged to allow for some measure of changing variable types without automatically breaking savegames. - Use these new methods to serialize the non-native variables in an object. This allows for achiving non-ints.
2016-04-03 04:10:43 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClass :: FindClass
//
// Find a type, passed the name as a name.
//
//==========================================================================
PClass *PClass::FindClass (FName zaname)
{
if (zaname == NAME_None)
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return static_cast<PClass *>(TypeTable.FindType(RUNTIME_CLASS(PClass), 0, zaname, nullptr));
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// PClass :: CreateNew
//
// Create a new object that this class represents
//
//==========================================================================
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
DObject *PClass::CreateNew()
Normalize line endings
2016-03-01 15:47:10 +00:00
{
BYTE *mem = (BYTE *)M_Malloc (Size);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert (mem != nullptr);
Normalize line endings
2016-03-01 15:47:10 +00:00
// Set this object's defaults before constructing it.
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (Defaults != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
memcpy (mem, Defaults, Size);
else
memset (mem, 0, Size);
- everything compiles and mostly works again.
2017-02-12 13:04:48 +00:00
if (ConstructNative == nullptr)
{
I_Error("Attempt to instantiate abstract class %s.", TypeName.GetChars());
}
Normalize line endings
2016-03-01 15:47:10 +00:00
ConstructNative (mem);
((DObject *)mem)->SetClass (const_cast<PClass *>(this));
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
InitializeSpecials(mem, Defaults, &PClass::SpecialInits);
Normalize line endings
2016-03-01 15:47:10 +00:00
return (DObject *)mem;
}
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
//
// PClass :: InitializeSpecials
//
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
// Initialize special fields (e.g. strings) of a newly-created instance.
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//
//==========================================================================
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
void PClass::InitializeSpecials(void *addr, void *defaults, TArray<FTypeAndOffset> PClass::*Inits)
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
// Once we reach a native class, we can stop going up the family tree,
// since native classes handle initialization natively.
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
if ((!bRuntimeClass && Inits == &PClass::SpecialInits) || ParentClass == nullptr)
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
return;
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
}
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
ParentClass->InitializeSpecials(addr, defaults, Inits);
for (auto tao : (this->*Inits))
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
- added property definitions to the ZScript parser. This will allow defining custom properties for the default block in custom base classes. See 'Health' for an example. - support transient object member variables for information that does not need to be put in a savegame. - fixed: special initialization of objects needs to pass the proper defaults along, otherwise the parent classes will use their own, inappropriate one.
2017-01-15 15:55:30 +00:00
tao.first->InitializeValue((char*)addr + tao.second, defaults == nullptr? nullptr : ((char*)defaults) + tao.second);
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
}
}
//==========================================================================
//
// PClass :: DestroySpecials
//
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
// Destroy special fields (e.g. strings) of an instance that is about to be
// deleted.
//
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
//==========================================================================
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
void PClass::DestroySpecials(void *addr)
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
// Once we reach a native class, we can stop going up the family tree,
// since native classes handle deinitialization natively.
if (!bRuntimeClass)
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
{
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
return;
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(ParentClass != nullptr);
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
ParentClass->DestroySpecials(addr);
Added code to initialize and destroy string variables in classes - Will require being able to add strings to non-native classes to actually test this.
2016-04-03 22:45:04 +00:00
for (auto tao : SpecialInits)
{
tao.first->DestroyValue((BYTE *)addr + tao.second);
}
}
Don't waste time looking for variables to (de)init in native classes
2016-04-04 19:57:43 +00:00
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClass :: Derive
//
// Copies inheritable values into the derived class and other miscellaneous setup.
//
//==========================================================================
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
void PClass::Derive(PClass *newclass, FName name)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
newclass->bRuntimeClass = true;
Normalize line endings
2016-03-01 15:47:10 +00:00
newclass->ParentClass = this;
newclass->ConstructNative = ConstructNative;
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
newclass->Symbols.SetParentTable(&this->Symbols);
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
newclass->TypeName = name;
newclass->mDescriptiveName.Format("Class<%s>", name.GetChars());
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
//==========================================================================
//
// PClassActor :: InitializeNativeDefaults
//
//==========================================================================
void PClass::InitializeDefaults()
{
- fixed: Non-actor classes never called InitializeDefaults to set up their special variables. - fixed: DThinkerIterator and DBlockThingsIterator did not have a default constructor that was safe to call from an out-of-game context.
2017-01-29 17:23:09 +00:00
if (IsKindOf(RUNTIME_CLASS(PClassActor)))
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(Defaults == nullptr);
- fixed: Non-actor classes never called InitializeDefaults to set up their special variables. - fixed: DThinkerIterator and DBlockThingsIterator did not have a default constructor that was safe to call from an out-of-game context.
2017-01-29 17:23:09 +00:00
Defaults = (BYTE *)M_Malloc(Size);
// run the constructor on the defaults to set the vtbl pointer which is needed to run class-aware functions on them.
// Temporarily setting bSerialOverride prevents linking into the thinker chains.
auto s = DThinker::bSerialOverride;
DThinker::bSerialOverride = true;
ConstructNative(Defaults);
DThinker::bSerialOverride = s;
// We must unlink the defaults from the class list because it's just a static block of data to the engine.
DObject *optr = (DObject*)Defaults;
GC::Root = optr->ObjNext;
optr->ObjNext = nullptr;
optr->SetClass(this);
// Copy the defaults from the parent but leave the DObject part alone because it contains important data.
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (ParentClass->Defaults != nullptr)
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
{
- fixed: Non-actor classes never called InitializeDefaults to set up their special variables. - fixed: DThinkerIterator and DBlockThingsIterator did not have a default constructor that was safe to call from an out-of-game context.
2017-01-29 17:23:09 +00:00
memcpy(Defaults + sizeof(DObject), ParentClass->Defaults + sizeof(DObject), ParentClass->Size - sizeof(DObject));
if (Size > ParentClass->Size)
{
memset(Defaults + ParentClass->Size, 0, Size - ParentClass->Size);
}
}
else
{
memset(Defaults + sizeof(DObject), 0, Size - sizeof(DObject));
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
}
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
assert(MetaSize >= ParentClass->MetaSize);
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
if (MetaSize != 0)
{
Meta = (BYTE*)M_Malloc(MetaSize);
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
// Copy the defaults from the parent but leave the DObject part alone because it contains important data.
if (ParentClass->Meta != nullptr)
{
memcpy(Meta, ParentClass->Meta, ParentClass->MetaSize);
if (MetaSize > ParentClass->MetaSize)
{
memset(Meta + ParentClass->MetaSize, 0, MetaSize - ParentClass->MetaSize);
}
}
else
{
memset(Meta, 0, MetaSize);
}
if (MetaSize > 0) memcpy(Meta, ParentClass->Meta, ParentClass->MetaSize);
else memset(Meta, 0, MetaSize);
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
}
Normalize line endings
2016-03-01 15:47:10 +00:00
}
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
if (bRuntimeClass)
{
// Copy parent values from the parent defaults.
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(ParentClass != nullptr);
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
if (Defaults != nullptr) ParentClass->InitializeSpecials(Defaults, ParentClass->Defaults, &PClass::SpecialInits);
if (Meta != nullptr) ParentClass->InitializeSpecials(Meta, ParentClass->Meta, &PClass::MetaInits);
- everything compiles and mostly works again.
2017-02-12 13:04:48 +00:00
for (const PField *field : Fields)
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
{
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
if (!(field->Flags & VARF_Native) && !(field->Flags & VARF_Meta))
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
{
- everything compiles and mostly works again.
2017-02-12 13:04:48 +00:00
field->Type->SetDefaultValue(Defaults, unsigned(field->Offset), &SpecialInits);
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
}
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
}
}
if (Meta != nullptr) ParentClass->InitializeSpecials(Meta, ParentClass->Meta, &PClass::MetaInits);
for (const PField *field : Fields)
{
if (!(field->Flags & VARF_Native) && (field->Flags & VARF_Meta))
{
field->Type->SetDefaultValue(Meta, unsigned(field->Offset), &MetaInits);
fixed and completed the special field init code. Strings can now be used as class members, and so can structs which contain strings. - made 'DamageMultiply' an actor property and moved the initialization of ConversationRoot to the property handler for the compiler to get this stuff out of the type classes. - consolidate default initialization into one function which performs all the required setup. The original implementation did this when adding the fields but that cannot work because at that time no defaults have been created yet. - fixed: When deriving a class the child class's defaults also must initialize the copied parent fields with special initialization. This part was completely missing. - removed DECORATE code for parsing native classes because it's no longer needed.
2016-11-11 13:40:32 +00:00
}
}
Normalize line endings
2016-03-01 15:47:10 +00:00
}
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
//==========================================================================
//
// PClass :: DeriveData
//
// Copies inheritable data to the child class.
//
//==========================================================================
void PClass::DeriveData(PClass *newclass)
{
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClass :: CreateDerivedClass
//
// Create a new class based on an existing class
//
//==========================================================================
PClass *PClass::CreateDerivedClass(FName name, unsigned int size)
{
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
assert(size >= Size);
Normalize line endings
2016-03-01 15:47:10 +00:00
PClass *type;
bool notnew;
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
const PClass *existclass = FindClass(name);
Normalize line endings
2016-03-01 15:47:10 +00:00
- restored some code in PClass::CreateDerivedClass that accidentally got removed.
2017-02-23 18:03:31 +00:00
if (existclass != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- restored some code in PClass::CreateDerivedClass that accidentally got removed.
2017-02-23 18:03:31 +00:00
// This is a placeholder so fill it in
if (existclass->Size == TentativeClass)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- restored some code in PClass::CreateDerivedClass that accidentally got removed.
2017-02-23 18:03:31 +00:00
type = const_cast<PClass*>(existclass);
if (!IsDescendantOf(type->ParentClass))
{
I_Error("%s must inherit from %s but doesn't.", name.GetChars(), type->ParentClass->TypeName.GetChars());
}
DPrintf(DMSG_SPAMMY, "Defining placeholder class %s\n", name.GetChars());
notnew = true;
}
else
{
// a different class with the same name already exists. Let the calling code deal with this.
return nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
}
else
{
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
type = static_cast<PClass *>(GetClass()->CreateNew());
Normalize line endings
2016-03-01 15:47:10 +00:00
notnew = false;
}
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
type->TypeName = name;
type->bRuntimeClass = true;
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
Derive(type, name);
- fixed: ZScript did not properly fill in tentatively created classes from defaults processing. The proper setup for such classes was only done in CreateDerivedClass, but not in FindClassTentative itself. This extends CreateDerivedClass to allow it to create a class without fully initializing it.
2016-11-12 12:09:19 +00:00
type->Size = size;
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
type->MetaSize = MetaSize;
- fixed: ZScript did not properly fill in tentatively created classes from defaults processing. The proper setup for such classes was only done in CreateDerivedClass, but not in FindClassTentative itself. This extends CreateDerivedClass to allow it to create a class without fully initializing it.
2016-11-12 12:09:19 +00:00
if (size != TentativeClass)
{
type->InitializeDefaults();
type->Virtuals = Virtuals;
DeriveData(type);
}
- restored ZDoom 2.8.1's version of FindClassTentative which just modified the existing class instead of replacing it and having to run a costly and volatile PointerSubstitution call to replace all pointers to the old one.
2017-02-08 18:52:33 +00:00
else
type->ObjectFlags &= OF_Transient;
Normalize line endings
2016-03-01 15:47:10 +00:00
if (!notnew)
{
type->InsertIntoHash();
}
return type;
}
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
//==========================================================================
//
// PStruct :: AddField
//
// Appends a new metadata field to the end of a struct. Returns either the new field
// or nullptr if a symbol by that name already exists.
//
//==========================================================================
PField *PClass::AddMetaField(FName name, PType *type, DWORD flags)
{
PField *field = new PField(name, type, flags);
// The new field is added to the end of this struct, alignment permitting.
field->Offset = (MetaSize + (type->Align - 1)) & ~(type->Align - 1);
// Enlarge this struct to enclose the new field.
MetaSize = unsigned(field->Offset + type->Size);
// This struct's alignment is the same as the largest alignment of any of
// its fields.
Align = MAX(Align, type->Align);
if (Symbols.AddSymbol(field) == nullptr)
{ // name is already in use
field->Destroy();
return nullptr;
}
Fields.Push(field);
return field;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
Add an AddField() override to PClass that extends the default instance
2016-04-03 21:18:59 +00:00
//
// PClass :: AddField
//
//==========================================================================
PField *PClass::AddField(FName name, PType *type, DWORD flags)
{
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
if (!(flags & VARF_Meta))
{
unsigned oldsize = Size;
PField *field = Super::AddField(name, type, flags);
- changed PClass:AddField so that it doesn't create defaults if none are present. The ZScript compiler will not create any defaults until after all fields have been added and this function created bad data if it was run on a class without defaults. - added checks for duplicate field names. - moved the tree node symbol tables out of PSymbolTreeNode to the worker data. That symbol is a bad location because it restricts the usefulness of the symbol class which is also needed for variables which use different AST structs. - fixed some memory management issues with the work classes for the compiler that became apparent after moving the symbol tables in there. In several places these were copied around, possibly losing data.
2016-10-09 23:18:47 +00:00
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
// Only initialize the defaults if they have already been created.
// For ZScript this is not the case, it will first define all fields before
// setting up any defaults for any class.
if (field != nullptr && !(flags & VARF_Native) && Defaults != nullptr)
{
Defaults = (BYTE *)M_Realloc(Defaults, Size);
memset(Defaults + oldsize, 0, Size - oldsize);
}
return field;
}
else
- changed PClass:AddField so that it doesn't create defaults if none are present. The ZScript compiler will not create any defaults until after all fields have been added and this function created bad data if it was run on a class without defaults. - added checks for duplicate field names. - moved the tree node symbol tables out of PSymbolTreeNode to the worker data. That symbol is a bad location because it restricts the usefulness of the symbol class which is also needed for variables which use different AST structs. - fixed some memory management issues with the work classes for the compiler that became apparent after moving the symbol tables in there. In several places these were copied around, possibly losing data.
2016-10-09 23:18:47 +00:00
{
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
unsigned oldsize = MetaSize;
PField *field = AddMetaField(name, type, flags);
// Only initialize the defaults if they have already been created.
// For ZScript this is not the case, it will first define all fields before
// setting up any defaults for any class.
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
if (field != nullptr && !(flags & VARF_Native))
- implemented the basics of a working metadata system. This will store class meta properties in a separate memory block so that it won't have to muck around with PClass - which made the implementation from the scripting branch relatively useless because extending the data wasn't particularly easy and also not well implemented. This can now be handled just like the defaults.
2017-02-27 22:28:19 +00:00
{
Meta = (BYTE *)M_Realloc(Meta, MetaSize);
memset(Meta + oldsize, 0, MetaSize - oldsize);
}
return field;
Add an AddField() override to PClass that extends the default instance
2016-04-03 21:18:59 +00:00
}
}
//==========================================================================
Normalize line endings
2016-03-01 15:47:10 +00:00
//
// PClass :: FindClassTentative
//
// Like FindClass but creates a placeholder if no class is found.
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
// This will be filled in when the actual class is constructed.
Normalize line endings
2016-03-01 15:47:10 +00:00
//
//==========================================================================
- fixed: ZScript did not properly fill in tentatively created classes from defaults processing. The proper setup for such classes was only done in CreateDerivedClass, but not in FindClassTentative itself. This extends CreateDerivedClass to allow it to create a class without fully initializing it.
2016-11-12 12:09:19 +00:00
PClass *PClass::FindClassTentative(FName name)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
if (name == NAME_None)
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
size_t bucket;
PType *found = TypeTable.FindType(RUNTIME_CLASS(PClass),
/*FIXME:Outer*/0, name, &bucket);
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (found != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
return static_cast<PClass *>(found);
}
PClass *type = static_cast<PClass *>(GetClass()->CreateNew());
- added multiple message levels for 'developer' CVAR so that the important stuff won't get drowned in pointless notification spam that's of no use to anyone. - made 'developer' CVAR persist across launches and added some menu entries for it. - added checks for 'developer' to ACS's CheckInventory function.
2016-08-28 07:55:04 +00:00
DPrintf(DMSG_SPAMMY, "Creating placeholder class %s : %s\n", name.GetChars(), TypeName.GetChars());
Normalize line endings
2016-03-01 15:47:10 +00:00
- properly handle all meta properties for inventory items.
2017-02-27 23:45:16 +00:00
assert(MetaSize == 0);
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
Derive(type, name);
Normalize line endings
2016-03-01 15:47:10 +00:00
type->Size = TentativeClass;
- fixed: Class and struct name lookup was not context aware. If a later module reused an existing name for a different class or struct type, this new name would completely shadow the old one, even in the base files. Changed it so that each compilation unit (i.e. each ZScript and DECORATE lump) get their own symbol table and can only see the symbol tables that got defined in lower numbered resource files so that later definitions do not pollute the available list of symbols when running the compiler backend and code generator - which happens after everything has been parsed. Another effect of this is that a mod that reuses the name of an internal global constant will only see its own constant, again reducing the risk of potential errors in case the internal definitions add some new values. Global constants are still discouraged from being used because what this does not and can not handle is the case that a mod defines a global constant with the same name as a class variable. In such a case the class variable will always take precedence for code inside that class. Note that the internal struct String had to be renamed for this because the stricter checks did not let the type String pass on the left side of a '.' anymore. - made PEnum inherit from PInt and not from PNamedType. The old inheritance broke nearly every check for integer compatibility in the compiler, so this hopefully leads to a working enum implementation.
2017-01-23 18:09:36 +00:00
TypeTable.AddType(type, RUNTIME_CLASS(PClass), 0, name, bucket);
Normalize line endings
2016-03-01 15:47:10 +00:00
return type;
}
- implemented handling for virtual function. Syntax-wise I chose to make it as strict as possible to reduce the chance of errors: Virtual base functions must be declared with the 'virtual' keyword, and overrides in child classes with the 'override' keyword. This way any mismatch in parameters that otherwise would cause silent failure will outright produce a compile error.
2016-11-11 19:05:07 +00:00
//==========================================================================
//
// PClass :: FindVirtualIndex
//
// Compares a prototype with the existing list of virtual functions
// and returns an index if something matching is found.
//
//==========================================================================
int PClass::FindVirtualIndex(FName name, PPrototype *proto)
{
for (unsigned i = 0; i < Virtuals.Size(); i++)
{
if (Virtuals[i]->Name == name)
{
auto vproto = Virtuals[i]->Proto;
if (vproto->ReturnTypes.Size() != proto->ReturnTypes.Size() ||
vproto->ArgumentTypes.Size() != proto->ArgumentTypes.Size())
{
continue; // number of parameters does not match, so it's incompatible
}
bool fail = false;
// The first argument is self and will mismatch so just skip it.
for (unsigned a = 1; a < proto->ArgumentTypes.Size(); a++)
{
if (proto->ArgumentTypes[a] != vproto->ArgumentTypes[a])
{
fail = true;
break;
}
}
if (fail) continue;
for (unsigned a = 0; a < proto->ReturnTypes.Size(); a++)
{
if (proto->ReturnTypes[a] != vproto->ReturnTypes[a])
{
fail = true;
break;
}
}
if (!fail) return i;
}
}
return -1;
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClass :: BuildFlatPointers
//
// Create the FlatPointers array, if it doesn't exist already.
// It comprises all the Pointers from superclasses plus this class's own
// Pointers. If this class does not define any new Pointers, then
// FlatPointers will be set to the same array as the super class.
//
//==========================================================================
void PClass::BuildFlatPointers ()
{
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (FlatPointers != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{ // Already built: Do nothing.
return;
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
else if (ParentClass == nullptr)
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
{ // No parent (i.e. DObject: FlatPointers is the same as Pointers.
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (Pointers == nullptr)
{ // No pointers: Make FlatPointers a harmless non-nullptr.
Normalize line endings
2016-03-01 15:47:10 +00:00
FlatPointers = &TheEnd;
}
else
{
FlatPointers = Pointers;
}
}
else
{
ParentClass->BuildFlatPointers ();
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
TArray<size_t> ScriptPointers;
// Collect all pointers in scripted fields. These are not part of the Pointers list.
for (auto field : Fields)
{
if (!(field->Flags & VARF_Native))
{
field->Type->SetPointer(Defaults, unsigned(field->Offset), &ScriptPointers);
}
}
if (Pointers == nullptr && ScriptPointers.Size() == 0)
Normalize line endings
2016-03-01 15:47:10 +00:00
{ // No new pointers: Just use the same FlatPointers as the parent.
FlatPointers = ParentClass->FlatPointers;
}
else
{ // New pointers: Create a new FlatPointers array and add them.
int numPointers, numSuperPointers;
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
if (Pointers != nullptr)
{
// Count pointers defined by this class.
for (numPointers = 0; Pointers[numPointers] != ~(size_t)0; numPointers++)
{
}
}
else numPointers = 0;
Normalize line endings
2016-03-01 15:47:10 +00:00
// Count pointers defined by superclasses.
for (numSuperPointers = 0; ParentClass->FlatPointers[numSuperPointers] != ~(size_t)0; numSuperPointers++)
{ }
// Concatenate them into a new array
- Turned DropItem into a plain struct again like it was before the scripting branch got merged. Making this an object had little to no advantage, except being able to remove the deleter code. Now, with some of the class data already being allocated in a memory arena so that freeing it is easier, this can also be used for the drop item lists which makes it unnecessary to subject them to the GC. This also merges the memory arenas for VM functions and flat pointers because both get deleted at the same time so they can share the same one.
2017-02-08 19:37:22 +00:00
size_t *flat = (size_t*)ClassDataAllocator.Alloc(sizeof(size_t) * (numPointers + numSuperPointers + ScriptPointers.Size() + 1));
Normalize line endings
2016-03-01 15:47:10 +00:00
if (numSuperPointers > 0)
{
memcpy (flat, ParentClass->FlatPointers, sizeof(size_t)*numSuperPointers);
}
- fixed: Script defined pointers were not added to the FlatPointers array.
2016-11-17 21:21:08 +00:00
if (numPointers > 0)
{
memcpy(flat + numSuperPointers, Pointers, sizeof(size_t)*numPointers);
}
if (ScriptPointers.Size() > 0)
{
memcpy(flat + numSuperPointers + numPointers, &ScriptPointers[0], sizeof(size_t) * ScriptPointers.Size());
}
flat[numSuperPointers + numPointers + ScriptPointers.Size()] = ~(size_t)0;
Normalize line endings
2016-03-01 15:47:10 +00:00
FlatPointers = flat;
}
}
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
//==========================================================================
//
// PClass :: BuildArrayPointers
//
// same as above, but creates a list to dynamic object arrays
//
//==========================================================================
void PClass::BuildArrayPointers()
{
if (ArrayPointers != nullptr)
{ // Already built: Do nothing.
return;
}
else if (ParentClass == nullptr)
{ // No parent (i.e. DObject: FlatPointers is the same as Pointers.
ArrayPointers = &TheEnd;
}
else
{
ParentClass->BuildArrayPointers();
TArray<size_t> ScriptPointers;
// Collect all arrays to pointers in scripted fields.
for (auto field : Fields)
{
if (!(field->Flags & VARF_Native))
{
field->Type->SetPointerArray(Defaults, unsigned(field->Offset), &ScriptPointers);
}
}
if (ScriptPointers.Size() == 0)
{ // No new pointers: Just use the same ArrayPointers as the parent.
ArrayPointers = ParentClass->ArrayPointers;
}
else
{ // New pointers: Create a new FlatPointers array and add them.
int numSuperPointers;
// Count pointers defined by superclasses.
for (numSuperPointers = 0; ParentClass->ArrayPointers[numSuperPointers] != ~(size_t)0; numSuperPointers++)
{
}
// Concatenate them into a new array
- Turned DropItem into a plain struct again like it was before the scripting branch got merged. Making this an object had little to no advantage, except being able to remove the deleter code. Now, with some of the class data already being allocated in a memory arena so that freeing it is easier, this can also be used for the drop item lists which makes it unnecessary to subject them to the GC. This also merges the memory arenas for VM functions and flat pointers because both get deleted at the same time so they can share the same one.
2017-02-08 19:37:22 +00:00
size_t *flat = (size_t*)ClassDataAllocator.Alloc(sizeof(size_t) * (numSuperPointers + ScriptPointers.Size() + 1));
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (numSuperPointers > 0)
{
memcpy(flat, ParentClass->ArrayPointers, sizeof(size_t)*numSuperPointers);
}
if (ScriptPointers.Size() > 0)
{
memcpy(flat + numSuperPointers, &ScriptPointers[0], sizeof(size_t) * ScriptPointers.Size());
}
flat[numSuperPointers + ScriptPointers.Size()] = ~(size_t)0;
ArrayPointers = flat;
}
}
}
Normalize line endings
2016-03-01 15:47:10 +00:00
//==========================================================================
//
// PClass :: NativeClass
//
- create struct types. - fixed class creation. There was an infinite loop and some missing checks for native classes. - do not write the compiler's symbols to the same symbol table as the output. The output must go to GlobalSymbols but the internal symbols must go to a namespace specific table that can be discarded after compilation.
2016-10-08 11:34:37 +00:00
// Finds the native type underlying this class.
Normalize line endings
2016-03-01 15:47:10 +00:00
//
//==========================================================================
const PClass *PClass::NativeClass() const
{
const PClass *cls = this;
while (cls && cls->bRuntimeClass)
cls = cls->ParentClass;
return cls;
}
- scriptified the Sigil. This isn't fully tested yet.
2016-11-28 13:39:25 +00:00
VMFunction *PClass::FindFunction(FName clsname, FName funcname)
{
- exported more parts of the joystick menus and also exported all strings for them to the string table. - exported the skeleton definition for list menus.
2017-02-13 16:45:03 +00:00
auto cls = PClass::FindClass(clsname);
- scriptified the Sigil. This isn't fully tested yet.
2016-11-28 13:39:25 +00:00
if (!cls) return nullptr;
auto func = dyn_cast<PFunction>(cls->Symbols.FindSymbol(funcname, true));
if (!func) return nullptr;
return func->Variants[0].Implementation;
}
- fixed: the restart CCMD must clear out all statically stored pointers to VM functions because they do not survive the shutdown that is needed before loading new data.
2017-01-31 12:41:23 +00:00
void PClass::FindFunction(VMFunction **pptr, FName clsname, FName funcname)
{
- exported more parts of the joystick menus and also exported all strings for them to the string table. - exported the skeleton definition for list menus.
2017-02-13 16:45:03 +00:00
auto cls = PClass::FindClass(clsname);
- fixed: the restart CCMD must clear out all statically stored pointers to VM functions because they do not survive the shutdown that is needed before loading new data.
2017-01-31 12:41:23 +00:00
if (!cls) return;
auto func = dyn_cast<PFunction>(cls->Symbols.FindSymbol(funcname, true));
if (!func) return;
*pptr = func->Variants[0].Implementation;
FunctionPtrList.Push(pptr);
}
- scriptified the Sigil. This isn't fully tested yet.
2016-11-28 13:39:25 +00:00
Normalize line endings
2016-03-01 15:47:10 +00:00
/* FTypeTable **************************************************************/
//==========================================================================
//
// FTypeTable :: FindType
//
//==========================================================================
PType *FTypeTable::FindType(PClass *metatype, intptr_t parm1, intptr_t parm2, size_t *bucketnum)
{
size_t bucket = Hash(metatype, parm1, parm2) % HASH_SIZE;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
if (bucketnum != nullptr)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
*bucketnum = bucket;
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
for (PType *type = TypeHash[bucket]; type != nullptr; type = type->HashNext)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
if (type->TypeTableType == metatype && type->IsMatch(parm1, parm2))
Normalize line endings
2016-03-01 15:47:10 +00:00
{
return type;
}
}
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
return nullptr;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// FTypeTable :: AddType - Fully Parameterized Version
//
//==========================================================================
void FTypeTable::AddType(PType *type, PClass *metatype, intptr_t parm1, intptr_t parm2, size_t bucket)
{
#ifdef _DEBUG
size_t bucketcheck;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(FindType(metatype, parm1, parm2, &bucketcheck) == nullptr && "Type must not be inserted more than once");
Normalize line endings
2016-03-01 15:47:10 +00:00
assert(bucketcheck == bucket && "Passed bucket was wrong");
#endif
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
type->TypeTableType = metatype;
Normalize line endings
2016-03-01 15:47:10 +00:00
type->HashNext = TypeHash[bucket];
TypeHash[bucket] = type;
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
type->Release();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// FTypeTable :: AddType - Simple Version
//
//==========================================================================
void FTypeTable::AddType(PType *type)
{
intptr_t parm1, parm2;
size_t bucket;
- moved the TypeTableType pointer from PClassType to PType. Removing this variable is needed to remove PClassType and PClassClass as the next step to eliminate all of PClass's subclasses in order to clean up the type system.
2017-02-07 19:25:52 +00:00
// Type table stuff id only needed to let all classes hash to the same group. For all other types this is pointless.
type->TypeTableType = type->GetClass();
PClass *metatype = type->TypeTableType;
Normalize line endings
2016-03-01 15:47:10 +00:00
type->GetTypeIDs(parm1, parm2);
bucket = Hash(metatype, parm1, parm2) % HASH_SIZE;
- implemented the backend for dynamic arrays. Still needs thorough testing but it should be complete. - use a memory arena to store flat pointers so that the messed up cleanup can be avoided by deallocating this in bulk. - added a new SO opcode to the VM to execute a write barrier. This is necessary for all objects that are not linked into one global table, i.e. everything except thinkers and class types. - always use the cheaper LOS opcode for reading pointers to classes and defaults because these cannot be destroyed during normal operation. - removed the pointless validation from String.Mid. If the values are read as unsigned the internal validation of FString::Mid will automatically ensure proper results.
2017-02-07 13:48:27 +00:00
assert(FindType(metatype, parm1, parm2, nullptr) == nullptr && "Type must not be inserted more than once");
Normalize line endings
2016-03-01 15:47:10 +00:00
type->HashNext = TypeHash[bucket];
TypeHash[bucket] = type;
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
type->Release();
Normalize line endings
2016-03-01 15:47:10 +00:00
}
//==========================================================================
//
// FTypeTable :: Hash STATIC
//
//==========================================================================
size_t FTypeTable::Hash(const PClass *p1, intptr_t p2, intptr_t p3)
{
size_t i1 = (size_t)p1;
// Swap the high and low halves of i1. The compiler should be smart enough
// to transform this into a ROR or ROL.
i1 = (i1 >> (sizeof(size_t)*4)) | (i1 << (sizeof(size_t)*4));
if (p1 != RUNTIME_CLASS(PPrototype))
{
size_t i2 = (size_t)p2;
size_t i3 = (size_t)p3;
return (~i1 ^ i2) + i3 * 961748927; // i3 is multiplied by a prime
}
else
{ // Prototypes need to hash the TArrays at p2 and p3
const TArray<PType *> *a2 = (const TArray<PType *> *)p2;
const TArray<PType *> *a3 = (const TArray<PType *> *)p3;
for (unsigned i = 0; i < a2->Size(); ++i)
{
i1 = (i1 * 961748927) + (size_t)((*a2)[i]);
}
for (unsigned i = 0; i < a3->Size(); ++i)
{
i1 = (i1 * 961748927) + (size_t)((*a3)[i]);
}
return i1;
}
}
//==========================================================================
//
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
// FTypeTable :: Clear
Normalize line endings
2016-03-01 15:47:10 +00:00
//
//==========================================================================
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
void FTypeTable::Clear()
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
for (size_t i = 0; i < countof(TypeTable.TypeHash); ++i)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
for (PType *ty = TypeTable.TypeHash[i]; ty != nullptr;)
Normalize line endings
2016-03-01 15:47:10 +00:00
{
- remove all type table entries from garbage collection. Like the symbols and the VM functions this is data that is static from startup until shutdown and has no need to be subjected to garbage collection. All things combined this reduces the amount of GC-sensitive objects at startup from 9600 to 600.
2017-02-08 21:43:20 +00:00
auto next = ty->HashNext;
delete ty;
ty = next;
Normalize line endings
2016-03-01 15:47:10 +00:00
}
}
memset(TypeHash, 0, sizeof(TypeHash));
}
#include "c_dispatch.h"
CCMD(typetable)
{
DumpTypeTable();
}
Reference in a new issue Copy permalink