zdoom/raze
0
0
Fork 0
mirror of https://github.com/ZDoom/Raze.git synced 2024-11-15 17:01:28 +00:00
Code Issues Projects Releases Packages Wiki Activity
16239 commits 34 branches 63 tags 171 MiB
Commit graph

16239 commits

This branch
This branch
All branches
Author SHA1 Message Date
Christoph Oelckers
f6d852b5a0 - removed the old section builder 2021-12-30 09:53:09 +01:00
Christoph Oelckers
b0e4b49771 - removed the node builder. 
With libtess2 working we do not need this anymore.
 2021-12-30 09:53:09 +01:00
Christoph Oelckers
bb9313454d - removed SectorGeometry class. 2021-12-30 09:53:09 +01:00
Christoph Oelckers
7a9596ae81 - use SectionGeometry for 3D rendering. 2021-12-30 09:53:09 +01:00
Christoph Oelckers
d8a3035bcc - invalidate plane meshes when the sector is retriangulated. 
After this any old mesh has become invalid and needs to be rebuilt.
 2021-12-30 09:53:09 +01:00
Christoph Oelckers
37e49ed775 After this any old mesh has become invalid and needs to be rebuilt. 
- use only one Section type.
 2021-12-30 09:53:09 +01:00
Christoph Oelckers
8e01d559e8 - store sector as an index in Section2 and renamed walls to lines. 
Now everything has the same name and type as in the old section.
 2021-12-30 09:53:08 +01:00
Christoph Oelckers
208ca4f783 - merge old SectionLine and new Section2Wall into one type and use the same global array. 2021-12-30 09:53:08 +01:00
Christoph Oelckers
d353787d31 - add a global list of section walls. 2021-12-30 09:53:08 +01:00
Christoph Oelckers
f8f6c345fc - use index fields in Section2Wall. 
There's little benefit of using pointers here - it inflates a critical render struct, does not make any code more readable and requires quite deep refactoring of hw_bunchdrawer.
Now, with all fields having the same tyoe and name a lot less code needs to be changed.
 2021-12-30 09:53:08 +01:00
Christoph Oelckers
a407e86ff6 - migrated automap to new SectionGeometry class. 2021-12-30 09:53:08 +01:00
Christoph Oelckers
362b824775 - ClipWindow WIP 2021-12-30 09:53:08 +01:00
Christoph Oelckers
4364e68d05 - automap work to transition to new sections. 2021-12-30 09:53:08 +01:00
Christoph Oelckers
b3873be514 - added some utilities to the Plane class that will be needed soon. 2021-12-30 09:53:07 +01:00
Christoph Oelckers
411e3e6f9c - make SectionGeometry create some data. 2021-12-30 09:53:07 +01:00
Christoph Oelckers
ca19c265f8 - interface from Section2 to the flat processing. 2021-12-30 09:53:07 +01:00
Christoph Oelckers
69593fd5c7 - Triangulation WIP 
# Conflicts:
#	source/core/sectorgeometry.cpp
 2021-12-30 09:53:07 +01:00
Christoph Oelckers
931903e291 - added libtess2 to the project. 2021-12-30 09:53:07 +01:00
Christoph Oelckers
449362dcb3 - better debug output. 2021-12-30 09:53:07 +01:00
Christoph Oelckers
a255e90c96 - more work on the new section creator. 2021-12-30 09:53:06 +01:00
Christoph Oelckers
b1bf12df9d - section fixes. 2021-12-30 09:53:06 +01:00
Christoph Oelckers
ad018db80a - new section builder. 2021-12-30 09:53:06 +01:00
Christoph Oelckers
9397eb6a19 - automatically handle one specific case of bad wall loops. 
Both RRRA E3L1.map and SW $yamato.map have this:

Wall x and Wall x+1 have identical properties and both reference Wall x+2 as point2.
The duplicate in this case can be safely deleted to make the sector well formed.
 2021-12-30 09:53:06 +01:00
Christoph Oelckers
a733de618f - new section WIP. 2021-12-30 09:53:06 +01:00
Christoph Oelckers
1bf654036a - delete duplicated walls in RRRA E1L3 and E1L6. 2021-12-30 09:48:50 +01:00
Christoph Oelckers
504bb08a64 - we do not really need the old rintersect function. 
Its only difference is using potentially overflowing 32 bit math for demo compatibility. Aside from that it has no real-life advantage.
 2021-12-26 23:10:07 +01:00
Christoph Oelckers
88b660a70d - removed compat.h entirely, all that was left was redundant #includes. 2021-12-26 23:10:07 +01:00
Christoph Oelckers
91cdd82c8a - removed all #include "compat.h" except the one in build.h 
Still all compiles fine.
 2021-12-26 23:10:07 +01:00
Christoph Oelckers
6db1ba47c2 - SW: fixed some savegame issues: 
* the saveable tables were initialized too late when reading in the actors.
* the abort messages for the saveables did not show. We need to use I_FatalError here, plus some actor list modifications to make this work.
 2021-12-26 23:10:07 +01:00
Christoph Oelckers
ae622274bd - SW: clear pointer holding variables in TerminateLevel. 
Some survive the level transition and would point to invalid data.
 2021-12-26 23:10:07 +01:00
Christoph Oelckers
a4ff7e5bd4 - SW: added some null checks to the SO interpolation code. 
It looks like this was able to access deleted sprites - by only checking for the index this never registered as a problem, but with pointers it does.
 2021-12-26 23:10:07 +01:00
Christoph Oelckers
a23d2ae656 - fixed OOB access in pushmove. 
The backwards search loop starts at one element behind the data it is supposed to check. If this is the last wall in the map it accesses undefined memory.

This tripped the range check in TArray for wall[].
 2021-12-26 23:10:06 +01:00
Mitchell Richters
4562a0ca97 - SW: Add some nullptr checks to some inlines to prevent noclip crashes. 2021-12-26 23:10:06 +01:00
Mitchell Richters
3b099e09cd - SW: Tidy some ‘newsect’ may be used uninitialized warnings from GCC. 2021-12-26 23:10:06 +01:00
Mitchell Richters
9e51f911ab - Blood: Clear writing to an object of type ‘class TObjPtr<Blood::DBloodActor*>’ with ‘private’ member ‘TObjPtr<Blood::DBloodActor*>::<anonymous>’ warning under GCC. 2021-12-26 23:10:06 +01:00
Mitchell Richters
2f567bdb74 - Exhumed: Fix two signed/unsigned comparison warnings in PlayFX2(). 
* Lines in question:
** 606 | if (chan->SoundID == nSound + 1)
** 621 | if (chan->SoundID == nSound + 1)
* All data sent to `PlayFX2()` and functions that wrap `PlayFX2()` are signed values, so keep signed all the way through.
 2021-12-26 23:10:06 +01:00
Mitchell Richters
a4d165a84e - Tidy some specified bound equals destination size warnings from GCC. 2021-12-26 23:10:05 +01:00
Christoph Oelckers
b89bba5ed1 - Duke: DDukeActor::setActiveCrane must also clear ownerActor. 
Apparently something in the crane code can cause a condition where this wasn't cleared before grabbing a player.
Formerly these two were the same variable but had to be separated.
 2021-12-26 23:10:05 +01:00
Christoph Oelckers
9e94573ae3 - free the current level before taking down other things on exit. 
Nor doing so can cause memory corruption when the object hierarchy gets deleted later which may still reference some stale data.
 2021-12-26 23:10:05 +01:00
Christoph Oelckers
d8ccfa1a5f - cleaned up nextsectorneighborz, added a safe variant and use this in all places where the sector pointer is not validated. 2021-12-26 23:10:05 +01:00
Mitchell Richters
18bd2e4d07 - SW: Destroy actor in KillActor() after clearing the pointed to sprite, not before it. 
* No idea if it matters but it seems logical that this is how it should be.
 2021-12-26 23:10:05 +01:00
Mitchell Richters
65a55d3589 - SW: Tidy up function declarations where return type was on its own line. 2021-12-26 23:10:05 +01:00
Mitchell Richters
1ee1c0c920 Revert "- fixed: updatesectorneighborz may not return -1." 
This reverts commit c074b0995648a4057c516e5646f5fe7a11719317.

* The commit notes talk about `updatesectorneighborz()`, but the change is applied to `nextsectorneighborz()`. Further to this, `nextsectorneighborz()` is only called from some ptr wrappers in `build.h` and they both test against `-1` return values.

```
inline sectortype* nextsectorneighborzptr(int16_t sectnum, int32_t refz, int16_t topbottom, int16_t direction)
{
	auto sect = nextsectorneighborz(sectnum, refz, topbottom, direction);
	return sect == -1? nullptr : &sector[sect];
}

inline sectortype* nextsectorneighborzptr(sectortype* sectp, int32_t refz, int16_t topbottom, int16_t direction)
{
	auto sect = nextsectorneighborz(sector.IndexOf(sectp), refz, topbottom, direction);
	return sect == -1? nullptr : &sector[sect];
}
```

* Also fixes broken Duke elevators and possibly a whole raft of issues.
 2021-12-26 23:10:04 +01:00
Mitchell Richters
953c7e4ed6 - Blood: Fix some collision issues found following refactoring. 
* Originating from 9a2ca3f392e29fe8df4167c271a7b04ab5e8959c "- use refactored Collision struct and clipmove in Blood, too.".
 2021-12-26 23:10:04 +01:00
Christoph Oelckers
7b14aacdf4 - match alignment of NPOT flat textures to the software renderer 
Tested in RR E1L4, sector 224+40, probably needs more checking in other places as well to confirm that it is correct there as well.
 2021-12-26 23:10:04 +01:00
Christoph Oelckers
b4a968841a - missing null check in Blood's map loader. 2021-12-26 23:10:04 +01:00
Christoph Oelckers
5e09a43e91 - added auto-patcher for overlapping wall ranges. 
This is for cases where the proper sector can be found by looking at the walls' point2 links.
 2021-12-26 23:10:04 +01:00
Christoph Oelckers
fee8df8a1b - fix range check for nextwall in map loader 2021-12-26 23:10:03 +01:00
Christoph Oelckers
177e84881e - added a warning for multiple sectors referencing the same walls. 
RR E3L1 contains such a setup, for example.
 2021-12-26 23:10:03 +01:00
Mitchell Richters
14b8eda436 - Get rid of some now irrelevant comments. 2021-12-26 23:10:03 +01:00