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.
- 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.
- Since the number of small allocations here is extremely high this will help a lot to prevent fragmentation and since most nodes are collected up front and this is done when no large resources are being loaded it won't cause heap spikes.
let Emit methods delete FxExpression arrays when they are done.
- For some reason the deletion process does not work 100%, there are always some nodes left behind and so far I haven't found them. This ensures that these arrays do not live any longer than needed.
- fixed: The state cast hack for DECORATE could not properly create state constants.
Instead they were passed to FxRuntimeStateIndex without resolving them to something constant. This adds proper handling of constant indices within that class.
* use the function build list instead of the function to pass the info. The function is permanent so not the best place for compile-time info.
* pass along the current state index which is needed to calculate the target state.
- made some tests about calling script code from native functions.
* scriptified A_SkullAttack to have something to test
* changed the A_SkullAttack call in A_PainShootSkull.
* use a macro to declare the function pointer. Using local static variable init directly results in hideous code for the need of being thread-safe (which, even if the engine was made multithreaded is not needed here.)
* Importsnt node here: Apparently passing an actor pointer to the VMValue constructor results in the void * version being called, not the DObject * version.
- Polygons will be clipped to bottomclip. If this is zero or below, they
will be clipped to the bottom of the screen instead. This keeps the
polygons from overwriting the status bar border for sofware 2D. The
hardware version ignores it, since it always draws the status bar border
every frame.
- fixed: The state index comparison against 0 was broken.
- fixed: Resolving codegen nodes must set the strictness flag per function so that ZSCRIPT and DECORATE are done properly.
- added an 'exact' parameter to FThinkerIterator's Next function. This is mainly for scripting which allows to do a lot more checks natively when running the iterator while looking for one specific class.
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.
- fixed: divisions wasted one register for each operation due to a double allocation.
- changed math operations to use less registers. There was a well-intended change to allocate the destination first, but the better approach is to first allocate the operands and free then before allocating the destination register.
- 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.
* Revert "Modify CMPJMP to produce more compact code (as far as VC++ is concerned, anyway)"
This reverts commit 6ff973a06b.
This modification did not work and broke the comparisons. Actually this had three problems:
* the asserts checked the wrong instruction
* the mask was not applied to regular comparisons.
* incrementing PC before testing does not work because 'test' references the PC.
* Revert "Modify CMPJMP to produce more compact code (as far as VC++ is concerned, anyway)"
This reverts commit 6ff973a06b.
This modification did not work and broke the comparisons. Actually this had three problems:
* the asserts checked the wrong instruction
* the mask was not applied to regular comparisons.
* incrementing PC before testing does not work because 'test' references the PC.
Daedalus triggers this with a 0x85 character which in Windows CP 1252 is the ellipsis (...) The converter will assume ISO-8859-1, though, but cannot do anything with these characters because they map to the font being used here.
- These require manual detection and overriding of the scaling factors to
0, because a right shift of (32-0) bits wraps around to 0 and results in
no shift at all rather than leaving the register zeroed out.
- dc_destorg is normally set to the upper-left corner of the view window.
If there is a border, then this won't coincide with the upper-left
corner of the screen, and DCanvas::FillSimplePoly would merrily write off
the end of the screen buffer.
- 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.
Added restriction of Require/Exclude to ZSDF (namespace = "ZDoom";).
A warning will be printed if a Require/Exclude block is detected in USDF (namespace = "Strife";).
Added two new sub-blocks for Choice blocks: Require and Exclude.
The syntax for both is the same as Cost blocks.
Require defines what item must be present in your inventory in order to show this choice/reply.
Exclude defines what item must not be present in your inventory in order to show this choice/reply.
If any Require/Exclude blocks are defined then this choice/reply will be hidden until all blocks of both types are satisfied.
- fixed: FxMinusSign trashed local variables that were used with negation.
- fixed: FxConditional only handled ints and floats, but not pointers and strings.
- fixed: A 'no states in non-actors' error was triggered, even for classes without any states.
- fixed several occurenced where vectors were treated as floats. NOTE: The entire codegen.cpp file needs to be carefully reviewed for bad use of the REGT_ constants, there's probably more places where using them has broken some type checks.
- fixed: committed test version of zscript.txt instead of changed actor.txt by accident.
Initialization and assignment for strings is working with this commit.
- removed the bogus optional value from the first A_Jump argument. A quick test with an older ZDoom revealed that this was never working - and implementing it would make things a lot more complicated, especially error checking in the code generator.
- fixed: The check for insufficient parameters to a function call was missing.
- removed mulv_kr and divv_kr instructions. The first are redundant and the second are useless. Maybe remove all other vector/const operations as well? They won't get used by the code generator.
- fixed disassembly of vector multiplication and division instructions.
- added initializer syntax for vectors. A vector can be set with vectorvar = (x,y,z); for a 3-dimensional vector and vectorvar = (x, y); for a 2-dimensional one.
- swapped parameters of two-parameter VelToAngle method, so that internal and script version are in line.
- fixed parameter asserts to handle NULL pointers properly.
- fixed: ZCCCompiler did not process array access nodes.
- fixed: Function argument names were not placed in the destination list by the compiler.
- scriptified several trivial functions from p_actionfunctions.cpp.
- disabled the assert in PType::GetRegType. This assert blocks any use to check for types that are incompatible with function parameters.
- pass the default parameter constants to the native functions. At the moment this is not used yet.
- use the function defaults to complete argument lists to script functions.
- fixed all default values that got flagged by the expression evaluator as non-constant. Most were state labels and colors which were defaulted to "". The proper value is null for states and 0 for colors.
- also replaced all "" defaults for names with "none".
This reverts commit 81449728d7.
Reverted because it compromises the IWAD file lookup and fixing it properly is not so trivial.
The skipping of adding the file name extension was not only broken, but even after fixing the code does not work if the IWADs are located outside the working directory.
This fixes an issue with DUMP 2 which looked for patches of the same name as the texture currently being defined and where the patches had the same use type as the composite texture. The function as implemented would only find the newly added composite and print an error.
- fixed: Script functions did not receive the function name when being created.
- relaxed the asserts for PARAM_STATE, because the VM knows nothing about ATAG_STATE. Any state variable's content (e.g. Actor.SeeState) will receive ATAG_GENERIC, rather than ATAG_STATE.
- added a 'NeedResult' flag so that certain operations can create shorter code if the result of the expression is not needed. So far only used for postdecrement/increment statements on local variables (which is the most frequent case where this matters.)
- fixed postincrement and decrement for local variables. Due to the result preservation semantics it created faulty code.
If done earlier they will not be able to detect overrides of sprites and graphics which are not part of the PATCHES lump. There was some fudging code to work around this problem but it was only partially working.
Now these textures only collect the texture name and use type during setup and resolve them after all textures have been created.
This was probably responsible for some weird behavior recently, but with the addition of the OF_Transient flag this outright crashed because it left NULL pointers on reload in places where they weren't checked for.
- 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.
- removed 'self' as a dedicated token. Internally this gets handled as a normal but implicitly named variable so the token just gets in the way of proper processing.
- removed P_ prefix from SpawnMissile export.
- fixed a crash with misnamed function exports.
- added the above for the 'if' condition. It works for integers, floats and pointers and will save 3 instructions if the condition is a non-boolean that can be implicitly casted to bool.
- allow class extensions.
These are separate blocks in different files that get concatenated to one class body for processing. The reason is to allow spreading the many functions in Actor over multiple files, so that they remain manageable. For example, all the Doom action functions should be in their respective files, but their symbols need to be in Actor. To extend a class, both files need to be in the same translation unit, so it won't allow user-side extension of internal classes.
- added a TESTN instruction. This is like TEST but negates the operand. This was added to avoid flooding the constant table with too many case labels. With TEST and TESTN combined, all numbers between -65535 and 65535 can be kept entirely inside the instruction. Numbers outside this range still use a BEQ instruction.
- This was only visible when using a screen wipe because the initial frame wiped
to would clamp the pitch to whatever undefined pitch range the player
had before the proper range was received.
- Instead of calculating lighting based from the left edge of the wall
segment the decal was on, it was calculated from the left edge of the
wall instead.
- added master and tracer to the list of exported variables.
- fixed: 'none' as class type must map to the real null pointer so that it won't get rejected by the stricter type checks.
- added handling for member function calls to zcc_compile.cpp.
- fixed: FxMemberFunctionCall may not delete the self expression if it gets passed on to the actual function call.
- fixed emission of the self pointer in FxVMFunctionCall. I did not realize that the self expression only sets up a register for the value, not pushing it onto the stack.
This will restrict them to the only classes that may use them: Weapon and CustomInventory.
Note: Should a mod surface which uses them improperly the better solution would be a warning message and NULLing the bogus code pointer instead of leaving them in Inventory.
Ironically this only requires a very minor change in the calling code and an added member for the VMFunction to tell that code how many parameters to pass.
This change will allow to turn the vast majority of action functions into regular members, the only ones that still need to be an action function are the few that actually use the pointers.
This is not testable right now because finally the action function mess has come full circle. The current setup makes it impossible to call action functions from non-action functions because the needed info is local to the functions.
Long avoided, this needs to be refactored now so that the different semantics for action functions are no longer needed.
- VC++ generated horribly stupid code for x64 when incrementing pc at the
beginning of each instruction by storing hundreds of copies of it for
every opcode executed. Incrementing pc at the end avoids this madness.
- It is possible I messed something up with this change. Hopefully not.
- fixed: FxMemberIdentifier checked for ClassPointers instead of object pointers to resolve the left hand side of the expression.
- allow comparison of pointers.
This also means that for now Lua-style multi-assignments are disabled, those should be easy to enable by making some changes to the assignment_statement grammar so that it doesn't recognize single assignments, but for now this is low priority because it adds a significant amount of complexity to do this right with functions that have multiple return values.
The generated object code can definitely use an optimization pass but that's something left for later when more things are working. Right now it creates one opcode more than necessary for all member accesses (instead of using the offset in the store command it calculates an actual address of the variable in another address register) and can create one too many for non-constant expressions being assigned to local variables (a move between two registers because the emitted expression on the right hand side does not know that it can emit to the actual variable's register.)
- create proper variable data from the function prototype instead of assuming that there's just 3 pointers.
- added a printable name to VMScriptFunction for error output during gameplay in case something goes wrong.
All break and continue statements were collected in one list, and each loop statement taking hold of that entire list, including the breaks and continues from previous outer loops.
Changed it so that loop statements contain the jump addresses themselves and set a pointer in FCompileContext, so that the jump point can be set directly in the loop statement - and an error printed if there is none in the resolving stage, not the emitting one.
Consolidated the identical backpatching code of the three loop statement nodes into a subfunction.
- disable generation of the parser's trace file in debug builds.
This increases parsing time by a factor of 15 and is only needed when debugging a problem in the grammar.
- handle all binary operators which are already implemented in the code generator.
- implemented sizeof/alignof operators in code generator.
- rewrote RequestAddress so that its return value is not the writability of an address but the mere existence. Also changed it to not output errors itself because those cannot be dealt with by the calling function.
- added the '>>>' (unsigned shift) operator. Although, with unsigned numbers available, this is technically not really needed, DECORATE supports this so ZScript should, too, if only for the benefit of making conversion tools easier to handle.
- fixed: FxBinary::ResolveLR' check for numeric operations was incomplete. Like far too many other places it just assumed that everything with ValueType->GetRegType() == REGT_INT is a numeric type, but for names this is not the case.
These were previously faked with the inverse plus a boolean not. Although this works, it either leads to sub-optimal code generation or some fudging to avoid the inefficient handling.
Just adding proper handling to the parser seems the easiest and most straightforward way to get around this. The code generator already can deal with these operations properly so there's no good reason to do it differently.
- 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.
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...
- implemented handling of the basic math operators so that heretic/beast.txt can be processed.
This is working, aside from still needing the type casts to properly transform the strings to class pointers.
- ANIMATED contained definitions for Doom, Heretic, and Strife, all
crammed into a single file. This meant that animations from one game
could erroneously make their way into maps for another game that
provided custom textures with names that matched textures that animated
in the other game. There are now three separate ANIMATED lumps with only
the animations defined for the original game and no others. The one
that gets loaded depends on the game being played.
- Added documentation for the ANIMATED file format to the comment for
FTextureManager::InitAnimated(), since I had to figure it out from the
code.
- Setting an actor's Crash state has the potential to destroy the actor if
the Crash state has one or more 0-tic states that end with Stop. This
was not taken into account when the object's Z velocity was 0, but it
was under the floor anyway.
For the random functions this class only handles the default-RNG version. The one with an explicit RNG needs to be done separately because the parser produces different output for them.
- added a truncation warning to FxIntCast, which only occurs with ZScript, not with DECORATE. FxBoolCast is intentionally left out because it would defeat the reason for this cast type.
- removed Self parameter from FxFunctionCall. Actual member function calls through an object require quite different handling so lumping these two together makes no sense.
- added a workaround to deal with ACS_NamedExecuteWithResult to both the compiler and FindClassMemberFunction. The way the ZScript compiler sets this up means that it will call the builtin, not the actual action function, so the parser needs to do some explicit check to get past the same-named action function.
- pass a proper self pointer to FxActionSpecial. Although it's still not being used, propagating design shortcuts through several function levels is a very, very bad idea.
* explicitly require passing the owning class when creating it.
* extract self pointer class when adding a variant.
* put the flags on the single variants, we can not fully rule out that they will be 100% identical, if variants ever get allowed.
* Allow PFunction to work without a VMFunction being attached.
* The Variant for a function must store the prototype itself instead of relying on the VMFunction it points to. Otherwise it would not be possible to reference a prototype during compilation of the function because it does not exist yet.
* Give the variant a list of the function's argument's names, because these are also needed to compile the function.
* create an anonymous function symbol when the function gets registered to the builder. At this point we have all the needed information to set it up correctly, but later this is no longer the case. This is the most convenient info to have here because it contains everything that's needed to compile the function in the proper context, so it has to be present when starting compilation.
* added some preparations to implement special handling for weapons and custom inventory items, which can run action functions in another actor's context. This part is not active yet but the basics are present in SetImplicitArgs.
- use FScriptPosition's error counter throughout the compiler so that there is only one counter for everything, not two.
Parts of the compiler use FScriptPosition, so this is easier to handle than having a separate counter in the compiler class. It also avoids having to pass the compiler object to any function where an error may be output. The TreeNodes contain sufficient data to be converted to an FScriptPosition and using that for error message formatting.
It looks like the setup of the action function and the adding of the states got inverted by something I tried, leaving the ActionFunc member empty on the real states.
- fixed a few problems that were encountered during conversion:
* action specials as action functions were not recognized by the parser.
* Player.StartItem could not be parsed.
* disabled the naming hack for PowerupType. ZScript, unlike DECORATE will never prepend 'Power' to the power's name, it always needs to specified by its full name.
* states and defaults were not checked for empty bodies.
* the scope qualifier for goto labels was not properly converted to a string, because it is an ENamedName, not an FName.
- Every update rolled into one, because I'm pretty sure I missed some while
updating lemon.c (not counting today's commits), since it wasn't always
updated at the same time as lemon.c.
- In particular, I think this check-in from 2016-06-06 was very important to
us after commit 3d5867d29e (For the
Lemon-generated parser, add a new action type SHIFTREDUCE and use it to
further compress the parser tables and improve parser performance.):
* Fix lempar.c so that the shift-reduce optimization works for error
processing.
- started with the AST converter. So far it only deals with direct function calls with simple constants as parameters.
- added an error condition for the defaults block to get rid of some asserts.
- fixed uninitialized counter variable in DECORATE parser.
- allow dottable_id of xxx.color so that the property parser can parse 'powerup.color'.
- fixed crash with actor replacement in script compiler.
- add the lump number to tree nodes because parts of the property parser need that to make decisions.
- removed test stuff.
- converted inventory.txt, player.txt and specialspot.txt to ZSCRIPT. These were the minimal files required to allow actor.txt to parse successfully.
- removed the converted files from the DECORATE include list so that these are entirely handled by ZSCRIPT now.
- split FinishActor into several functions. While DECORATE can, ZSCRIPT cannot do all this in one go.
- split the state finalization into several class-specific virtual functions.
* everything related to scripting is now placed in a subdirectory 'scripting', which itself is separated into DECORATE, ZSCRIPT, the VM and code generation.
* a few items have been moved to different headers so that the DECORATE parser definitions can mostly be kept local. The only exception at the moment is the flags interface on which 3 source files depend.
To do the rest, some cleanup is needed first, to untangle the DECORATE parser from the actual code generation so that the low end stuff can actually be reused here instead of having to be redone.
Ultimately, thingdef should only contain code that is directly related to the DECORATE parser, but that's not the case with this file. It's only function definitions which get used during gameplay and will also be accessed by ZScript.
The change is intentionally on master so that pull requests can adjust to it now instead of creating conflicts later.
* Goto did not support the class scope operator '::'. Like in DECORATE, this cannot be done with a simple '.' because it creates semantic problems with first part of a state label. This requires different syntax so that it can unambiguously distinguish between a scope identifier and the actual label
* Goto used the incorrect token PLUS for '+' instead of ADD.
* The state's duration was not stored in the AST.
* Truncating the sprite name inside the parser is probably not the best idea because it used a simple Printf to report this. Let's do this during processing of the AST where this can be properly handled as an error.
- Sets the absolute amount of an inventory actor.
- Limits itself to the range [0, MaxAmount]. Setting beyondMax to true disregards the MaxAmount. Default is false.
- switched the types of the internal 'self' and 'stateowner' parameters so that they get assigned correctly. I can't tell if this will error out if fields get accessed from the caller with the wrong class, but for actual scripting to work these must be correct.
The committed 'actor.txt' can be parsed successfully, with the exception of a few subclass references that cannot be resolved yet.
This adds:
* builtin types color, state and sound.
* ending a parameter list with an ellipsis to declare a varargs list. (A_Jump uses this.)
* allowing to declare optional arguments by giving them a default value.
* adding an 'action' qualifier for function declarations.
This uses the same property and flag tables as DECORATE with a few changes:
* it sets the parse mode to strict, so that several DECORATE warnings are now errors.
* trying to change a deprecated flag will print a warning.
* setting of editor numbers, spawn and conversation ID id not possible. Use MAPINFO to do this.
* all subclass flags must use the qualified name now (e.g. +ALWAYSPICKUP will print an error.)
* the scriptable Damage property is not yet implemented. This will require a special case with a differently named property in the processing function because in the AST it is no longer possible to distinguish between a damage value and a constant damage function.
This had been defined as a regular compound statement but in the context this will be used in, that makes very little sense, because all it can do is set some constant values.
The most important thing here is that it doesn't provide an unnecessary learning curve to its users and doing it this way will not only ensure that but also avoid redundant documentation.
To allow initialization of other user-defined properties it will require some extensions but that's a job for later and can just as easily be done in the current framework, rather than throwing everything out and start from zero.
- 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.
- fixed: The tree nodes for classes and struct members were stored in the global tree nodes table.
- sort variable declarations into their own list for processing.
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.
- 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.
This is the first thing the compiler has to do to get access to the class's symbol table. Of course at this point the final size of a class is not known yet so these are currently all treated as tentative.
- place generated symbols into GlobalSymbols instead of a scratch table that will be discarded right away.
- allow the state object to change source file scanners (I hope this works, but the old implementation was unable to do more than one with with a parse state so I had to change it.)
- It can now parse constants.txt and insert everything in it into the global symbol table and make subsequent DECORATE compile properly.
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.
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.
There have been reports about crashes in here with Linux that point to some of the code that gets called here doing unwanted things on the owner, so with these links cleared that should no longer be possible.
- This was fine with fixed point numbers, since they could never be
outside of short range when converted to regular ints. With floating
point numbers now, that condition no longer holds.
- allow recursive linking of $random definitions (as long as they do not link back, see above.)
- fixed the sound precaching which did not handle $alias inside $random. Normally this went undetected but in cases where the random sound index was the same as a sound index in the current link chain this could hang the function.
