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
- 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.
I wish I had realized this the last time it came up - it would have saved me a lot of trouble.
But as it turns out, the more recent travelling code makes all of this completely unnecessary, working perfectly fine with deleting the player pawns along with the rest of the thinkers before loading the stored ones from the savegame (and getting rid of those in G_FinishTravel.)
And with a sane savegame format that does not depend on side effects from how the thinker serializing handled linking into the lists the old code was even harmful, leaving voodoo dolls behind.
I had the exact same effect when I tried to reshuffle some things for reliably restoring portals, but did not make the connection to interference between two mutually incompatible player travelling mechanisms that just worked by sheer happenstance with the original order of things.
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.
- always make the top level object randomaccess when opening a JSON file for reading. Some things won't work right if this is opened for sequential access.
Since this list is excluded from regular thinker cleaning, anything that may survive through the end of G_FinishTravel will endlessly multiply and severely break the following savegames or just simply crash on broken pointers.
Note that this required splitting P_SerializeWorld, because sector_t and FSectorPortal contain some actor pointers, for which the same rule applies: Portal linking can only be done after all sectors have been read, meaning it cannot be done along with the rest of the data in these structures.
Obviously such a change breaks savegame compatibility so the min. savegame version had to be increased again.
Two issues:
Portal linking requires all skyboxes in the sectors to be present, without them some info is not there when needed.
UpdateWaterLevel was called in AActor::Serialize, which operated on the freshly loaded level where lots of things haven't been set up yet and plane heights may be wrong.
become ungodly slow when using mods with complex DECORATE. The GCC debug builds run just
fine, however. Hopefully this is something that can be fixed later with an assembly-optimized
version of the main VM loop, because I don't relish the thought of being stuck with GDB
for debugging.)
- Fixed: The ACS_Named* action specials were erroneously defined as taking strings instead of
names.
- Fixed: Copy-paste error caused FxMultiNameState::Emit to generate code that called
DecoNameToClass instead of DecoFindMultiNameState.
- Updated FxActionSpecialCall::Emit for named script specials.
- Fixed inverted asserts for FxMinusSign::Emit and FxUnaryNotBitwise::Emit.
SVN r3893 (scripting)
because it exposed a design flaw in the thinker system:
Having every single actor default to the highest available statnum means that
nothing can be placed in a slot where it is guaranteed to be run after all actors
have ticked. But this is required for any thinker that moves an actor
(i.e. AActorMover and DSectorEffect.) With DSectorEffect it just went unnoticed
because they were added at the end of the list so almost nothing they moved was
behind them in a thinker list. However, when an actor was spawned on a moving
floor it did not move smoothly. The default statnum is now 100 so that there's
sufficient slots above where such thinkers can be placed.
SVN r2060 (trunk)
- AActor::PreExplode is gone now that the last item that was using it has been converted.
- Converted the Sigil and the remaining things in a_strifeitems.cpp to DECORATE.
SVN r1132 (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)
way wasn't playing well with the write barriers.
- Fixed: DFrameBuffer::WriteSavePic needs to fix the canvas in place while
using it.
- Fixed: The sound system was updated every frame, which is a complete waste
of time. Doing it only once each tic is quite sufficient, since nothing
really moves between tics, even if the display makes it look otherwise.
SVN r799 (trunk)
FreshThinkers lists. These thinkers would still be saved in the savegame if
there were other references to them, but they would not go in any thinker
lists when loading the savegame. An easy way to exercise this bug is to
give yourself something new with the console and then save the game right
away without closing the console by using the "save" command. The item
would be stored in the savegame thanks to its presence in the Inventory
list, but it wouldn't be in a thinker list when the game was reloaded
because it was still in the FreshThinkers list when the game was saved.
SVN r763 (trunk)
it was previously declared unrecommended and deprecated.
- Added Blzut3's sbarinfo update #10:
* Fixed: drawmugshot now adjust to the player's max health like it should.
* drawgem can now accept negative paddings.
* Added: You can now specify a border on drawbar which will reserve x
amount of pixels on each side of the fg image.
* Added: weaponammo command which allows for determining what kind of ammo
the current weapon uses.
* Added: Drawkeybar to SBarInfo
SVN r730 (trunk)
- Added DECORATE support for setting an actor's args. If this is done
it will also disable the actor's special that can be set in a map. This
is for actors that want to use A_CountdownArg or similar functions
that use the args for something different than the special's parameters.
- Converted a_sharedmisc.cpp to DECORATE.
- Added a new NActorIterator that can search for classes specified by name.
- Added a new constructor to TThinkerIterator that can search
for DECORATE defined subclasses that are not represented by a real C++ class.
- Fixed: BuildInfoDefaults must set the parent symbol table so that
all actors can get to the global symbols stored in AActor.
- Fixed some minor inconsistencies in the Arch-Vile's DECORATE definition.
- Fixed: A_VileAttack moved the flame without relinking it into the sector
lists. It also forgot to set the z-position correctly. (original Doom bug.)
- Fixed: The Doom 2 cast finale didn't work with the dynamic state name handling.
SVN r401 (trunk)
- Changed: Decal scales now use full precision fixed point numbers.
- Changed: Keeping impact decals in their own statlist is enough to keep track
of them for when one needs to be destroyed. There's no need to maintain a
separate list for them.
- Fixed: Decal actors did not spread their decals across neighboring walls.
- Fixed: Decal groups did not initialize their IDs and could not be reliably
used with the decal actor.
- Fixed: Decals on moving polyobjects were not interpolated. R_RenderDecal()
now uses the decal's LeftDistance to calculate its visible location, so it
always stays in sync with the wall's vertices. This also lets me dump some
code from the polyobjects that maintained the decals' (x, y) coordinates.
Also, the decals' x and y information is redundant and can be removed.
Doing this revealed a bug with slider decals and horizontal sliding:
That is, it didn't work at all. I have opted to simply remove the horizontal
sliding support so that I don't have to worry about what happens when a
decal slides across the edge of a wall.
- Fixed: DBaseDecal::LeftDistance was calculated as a 30.2 fixed point number.
It should be 2.30 fixed point.
SVN r35 (trunk)
