This addresses a very strange crash I encounteded while travelling in a hub, and ended up with a NULL pointer after the 'Serialize' call which means that some code cleared the variable that is currently being deserialized. I was completely unable to find out what caused this because there is so much recursion going on in the deserializer. All actions on the deserialized actor are now being done with a local copy of that variable so that altering the actual one won't have any adverse effects.
Since decals may have thinkers attached this will crash when such a savegame gets loaded, because the thinker lists get reset in P_SerializeThinkers, deleting any thinker that already was processed.
I also added an error message that immediately aborts the save process if such an out-of-sequence thinker is attempted to be written out.
This obviously breaks savegame compatibility again...
- 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.
- Converted P_MovePlayer and all associated variables to floating point because this wasn't working well with a mixture between float and fixed.
Like the angle commit this has just been patched up to compile, the bulk of work is yet to be done.
Patched up everything so that it compiles without errors again. This only addresses code related to some compile error. A large portion of the angle code still uses angle_t and converts back and forth.
the DObject serialization template. Therefore, FArchive::SerializeObject() needs to be
able to handle them specially from other objects.
SVN r3924 (scripting)
- Fixed: Don't access class metadata at all in DObject::PropagateMark if the type system is shutdown.
- Fixed: If FCompressedMemFile::Reopen() fails, then it would try to double-free memory when deleted.
SVN r3688 (trunk)
instead of PClass::m_Types (now PClass::AllClasses).
- Removed ClassIndex from PClass. It was only needed by FArchive, and maps take care of the
problem just as well.
- Moved PClass into a larger type system (which is likely to change some/lots once I try and actually use it and have a better feel for what I need from it).
SVN r2281 (scripting)
A_SetUserVar/SetUserVariable/GetUserVariable now take a variable name
instead of an array index. A_SetUserArray/SetUserArray/GetUserArray
have been added to access elements in user-defined arrays.
SVN r1933 (trunk)
issues that caused its inclusion. Is an optimized GCC build any faster
for being able to use strict aliasing rules? I dunno. It's still slower
than a VC++ build.
I did run into two cases where TAutoSegIterator caused intractable problems
with breaking strict aliasing rules, so I removed the templating from it,
and the caller is now responsible for casting the probe value from void *.
- Removed #include "autosegs.h" from several files that did not need it
(in particular, dobject.h when not compiling with VC++).
SVN r1743 (trunk)
This seems to fix crashes on GCC 64-bit builds when saving games. Not sure
if it was a GCC bug or my bug, since it worked fine with VC++, but since the
code that calculates the index only returns a DWORD, storing it as a size_t
was rather pointless.
SVN r969 (trunk)
list, it should just be ignored during the propagation stage.
- After sleeping on it and realizing what was really going in, I generalized
the inventory fix from the 13th: The actor is flagged by Destroy(), then it
is later inserted into the thinker list by DThinker::SerializeAll(). So
rather than unlinking the skipped player from their inventory, just make
sure any flagged thinkers aren't inserted into a list.
- Fixed: FCanvasTextureInfo::Viewpoint needed a read barrier, and the whole
list should serve as a root.
- Reimplemented SPC playback as a custom codec for FMOD.
- Removed spc_frequency, because snes_spc only supports the SPC's native
frequency of 32000 Hz.
SVN r806 (trunk)
copies of the players stored in the archive would be destroyed and break
apart the STAT_INVENTORY thinker list. Now they aren't destroyed while
loading the archive but are unlinked from the inventory list instead, so
the garbage collector will get them later.
- Changed pointer substitution to return the number of pointers changed.
SVN r803 (trunk)
- Fixed: The MusicVolumes list was allocated with M_Malloc but freed with
delete.
- Fixed: demobuffer was inconsistantly handled with new[]/delete[] and
malloc/free.
- Added used memory tracking to M_Malloc() and M_Realloc(). This
necessitated the addition of an M_Free() call to track frees.
- Removed M_Calloc since it was only used in one place, and can just as well
be done with an M_Malloc/memset pair.
- Bumped DEMOGAMEVERSION for the new net controller codes.
SVN r751 (trunk)
- The stat meters now return an FString instead of sprintfing into a fixed
output buffer.
- NOASM is now automatically defined when compiling for a non-x86 target.
- Some changes have been made to the integral types in doomtype.h:
- For consistancy with the other integral types, byte is no longer a
synonym for BYTE.
- Most uses of BOOL have been change to the standard C++ bool type. Those
that weren't were changed to INTBOOL to indicate they may contain values
other than 0 or 1 but are still used as a boolean.
- Compiler-provided types with explicit bit sizes are now used. In
particular, DWORD is no longer a long so it will work with both 64-bit
Windows and Linux.
- Since some files need to include Windows headers, uint32 is a synonym
for the non-Windows version of DWORD.
- Removed d_textur.h. The pic_t struct it defined was used nowhere, and that
was all it contained.
SVN r326 (trunk)
while and killing centaurs with the flechette.
- Fixed: Moving to an old level in a hub caused the old player's inventory to
spawn owned by the current player (but still hanging off the old player), so
the game would hang when trying to delete it.
- Modified re2c so that it doesn't add a date to the file it generates. Thus,
if it regenerates a file during a full rebuild, SVN won't see it as a change.
Also updated it to 0.10.5.
- Fixed: SC_GetString() did not properly terminate sc_String when the last
token in the file had no white space after it. Since I could not actually
find the problem (it works fine in debug mode and I saw no logic errors),
I decided to take this opportunity to reimplement it using an re2c-generated
scanner. Now it's 1.6x faster than before and correctness is easier to
verify.
- Fixed: FMODSoundRenderer::Shutdown() also needs to reset NumChannels.
- Added back the Manifest to zdoom.rc for non-VC8 Windows compilers.
- Fixed MinGW compilation again. Now it uses the same method as Makefile.linux
to find all the source files so that it doesn't need to be manually updated
each time source files are added or removed.
- Added the SVN revision number to the version string. A new tool is used to
obtain this information from the svnversion command and write it into a
header file. If you don't have the svn command line tools installed or didn't
check it out from the repository, you can still build. I added some rules for
this to Makefile.linux, and I assume they work because they do for
Makefile.mingw.
- Fixed: MIDISong2 did not delete MusHeader in its destructor.
SVN r200 (trunk)
data enough (2 bits instead of 8), so it was super loud and aliased.
- Fixes for GCC 4.1: Several type-punned pointer warnings, but more
importantly, declaring a friend function inside a class body is no longer
enough to declare that function globally; you must declare it again outside
the class.
- Upgraded FArchive::SerializePointer so that it can store 32-bit indices.
- ACS printing pcodes now build their string in an FSttring instead of a fixed
sized buffer on the stack.
SVN r145 (trunk)
memcpy to copy the player structures.
- Fixed compilation with MinGW again and removed most of the new warnings.
And following is the log that I forgot to paste in for the previous commit:
- Changed the memory management for FString. Instead of using a garbage
collected heap, it now uses normal heap calls and reference counting to
implement lazy copying. You may now use bitwise operators to move
(but not copy!) FStrings around in memory. This means that the
CopyForTArray template function is gone, since TArrays can now freely
move their contents around without bothering with their specifics.
There is one important caveat, however. It is not acceptable to blindly 0
an FString's contents. This necessitated the creation of a proper
constructor for player_s so that it can be reset without using memset. I
did a quick scan of all memsets in the source and didn't see anything else
with a similar problem, but it's possible I missed something.
- Fixed: Build tiles were never deallocated.
- Fixed: Using Build's palette.dat only got half the palette right.
SVN r117 (trunk)
functions added to the exit chain with atterm so that they can be called
in a deterministic order and not whatever order the linker decides to put
them in.
- Fixed: DCajunMaster did not free its getspawned.
- Fixed: P_FreeLevelData() did not free ACS scripts.
- Fixed: Level snapshots were not freed at exit.
- Fixed: The save/load menu list was not freed at exit.
- Fixed: FCompressedMemFile needs a destructor to free the m_ImplodedBuffer.
- Fixed: G_DoLoadGame() did not free the engine string.
- Fixed: M_ReadSaveStrings() did not free the engine string.
- Fixed: Processing DEM_SAVEGAME did not free the pathname string.
- Added a check for truncated flats to FFlatTexture::MakeTexture() because
Heretic's F_SKY1 is only four bytes long.
- Added a dump of the offending state to the "Cannot find state..." diagnostic.
- Fixed: FCompressedFile did not initialize m_Mode in its default constructor.
- Fixed: Heretic and Hexen status bars did not initialize ArtiRefresh.
- Fixed: PNGHandle destructor should use delete[] to free TextChunks.
SVN r111 (trunk)
merge a lot of these static destructor-only structs into regular
functions added to the exit chain with atterm so that they can be called
in a deterministic order and not whatever order the linker decides to put
them in. (Interestingly, the amount of memory used when repeatedly
executing the same map command at the console varies up and down, but it
now stays relatively stable rather than increasing unbounded.)
- Fixed: The list of resolutions in the video modes menu was not freed
at exit.
- Fixed: mus_playing.name was not freed at exit.
- Fixed: SN_StopAllSequences() should be called at the start of
P_FreeLevelData(), not just before the call to P_SetupLevel() in
G_DoLoadLevel(), so it can run even at exit. And C_FullConsole() can
call P_FreeLevelData() to free more memory too.
- Fixed: StatusBar was not freed at exit.
- Fixed: spritesorter was not freed at exit.
- Fixed: Bad things happened if FString's data pool was destroyed before
all C_RemoveTabCommand() calls were made.
- Added an overload for FArchive << FString.
- Fixed: The players' log text was not freed at exit.
- Fixed: Bot information was not freed at exit.
- Fixed: doomcom was not freed at exit. But since it's always created,
there's no reason why it needs to be allocated from the heap. My guess
is that in the DOS days, the external packet driver was responsible for
allocating doomcom and passed its location with the -net parameter.
- Fixed: FBlockNodes were not freed at exit.
- Fixed: Openings were not freed at exit.
- Fixed: Drawsegs were not freed at exit.
- Fixed: Vissprites were not freed at exit.
- Fixed: Console command history was not freed at exit.
- Fixed: Visplanes were not freed at exit.
- Fixed: Call P_FreeLevelData() at exit.
- Fixed: Channel, SoundCurve, and PlayList in s_sound.cpp were not freed at
exit.
- Fixed: Sound sequences were not freed at exit.
- Fixed: DSeqNode::Serialize() did not resize the m_SequenceChoices array
when loading.
SVN r106 (trunk)
without any explicit constructors being called.
- Fixed: ZTwinedTorchUnlit, ZWallTorchUnlit, ZFireBullUnlit, and ZCauldronUnlit
were missing game filters.
SVN r79 (trunk)
- Fixed some functions that were declared as taking size_t's but defined as taking
unsigned ints.
- Added a dummy object to delete sound environments on exit.
- Fixed: FWarpTexture did not delete its Spans when destroyed.
- Changed wadclusterinfos and wadlevelinfos arrays into TArrays.
- Added the TypeInfo::AutoTypeInfoPtr for TypeInfo::m_RuntimeActors so they can
be automatically deleted.
- Changed TypeInfo::m_Types into a TArray so it will be automatically deleted
on exit.
- Fixed: TArray::Resize() did not deconstruct entries when shrinking the array.
- Changed TArray::Push() so that it calls Grow() instead of duplicating the growth
calculations itself.
- Calling TArray::Grow() for a small amount when the array is short should grow it
a bit more than it was doing.
SVN r76 (trunk)
actually work.
- Turned the list of TabCommands into a TArray because I saw lots of console
commands in the memory leak report at exit. Then I realized those were actually
key bindings, so I changed the Bindings and DoubleBindings arrays into FString
arrays.
- Fixed: FStringCVar was missing a destructor.
- Added TArray::Insert().
- Fixed: TArray::Delete() used memmove().
- Renamed Malloc(), Realloc(), and Calloc() to M_Malloc(), M_Realloc(), and
M_Calloc() so that the debug versions can be defined as macros.
- Enabled the CRT's memory leak detection in WinMain().
- Moved contents of PO_DeInit() into P_FreeLevelData().
- Removed "PolyBlockMap = NULL;" from P_SetupLevel(), because the P_FreeLevelData()
call it makes next does the exact same thing, but also freeing it if needed.
- Fixed: Unneeded memcpy in UnpackUserCmd() when ucmd and basis are the same
SVN r75 (trunk)
its tiles are loaded from the same directory.
- RFF files now load their entire directories into the lumplist.
- Added char * and const char * type coversions for FString, so FStrings can be
freely passed to functions expecting C strings. (Except varargs functions,
which still require manually fetching the C string out of it.)
- Renamed the name class to FName.
- Renamed the string class to FString to emphasize that it is not std::string.
SVN r74 (trunk)
- Added multiple-choice sound sequences. These overcome one of the major
deficiences of the Hexen-inherited SNDSEQ system while still being Hexen
compatible: Custom door sounds can now use different opening and closing
sequences, for both normal and blazing speeds.
- Added a serializer for TArray.
- Added a countof macro to doomtype.h. See the1's blog to find out why
it's implemented the way it is.
<http://blogs.msdn.com/the1/articles/210011.aspx>
- Added a new method to FRandom for getting random numbers larger than 255,
which lets me:
- Fixed: SNDSEQ delayrand commands could delay for no more than 255 tics.
- Fixed: If you're going to have sector_t.SoundTarget, then they need to
be included in the pointer cleanup scans.
- Ported back newer name code from 2.1.
- Fixed: Using -warp with only one parameter in Doom and Heretic to
select a map on episode 1 no longer worked.
- New: Loading a multiplayer save now restores the players based on
their names rather than on their connection order. Using connection
order was sensible when -net was the only way to start a network game,
but with -host/-join, it's not so nice. Also, if there aren't enough
players in the save, then the extra players will be spawned normally,
so you can continue a saved game with more players than you started it
with.
- Added some new SNDSEQ commands to make it possible to define Heretic's
ambient sounds in SNDSEQ: volumerel, volumerand, slot, randomsequence,
delayonce, and restart. With these, it is basically possible to obsolete
all of the $ambient SNDINFO commands.
- Fixed: Sound sequences would only execute one command each time they were
ticked.
- Fixed: No bounds checking was done on the volume sound sequences played at.
- Fixed: The tic parameter to playloop was useless and caused it to
act like a redundant playrepeat. I have removed all the logic that
caused playloop to play repeating sounds, and now it acts like an
infinite sequence of play/delay commands until the sequence is
stopped.
- Fixed: Sound sequences were ticked every frame, not every tic, so all
the delay commands were timed incorrectly and varied depending on your
framerate. Since this is useful for restarting looping sounds that got
cut off, I have not changed this. Instead, the delay commands now
record the tic when execution should resume, not the number of tics
left to delay.
SVN r57 (trunk)
