Doomsday Engine

The crash was caused by null pointer access. Some of the particles created by the "lightning" hit effect didn't have a current sector set, presumably because they were being spawned outside the map.

The patch from git does not fixes SEGV. The bug can be reproduced as described in the fitst message.

Some offline comments:
- - - - - - - - - - - -

This assert fails in P_NewParticle():

pt->sector = R_PointInSubsector(FIX2FLT, FIX2FLT)->sector;
assert(pt->sector);

danij: Hmm, the would indeed suggest a degenerate nodebuild case. I'll test this once I've merged the map-cache changes (I've addressed the two causes of degenerates known to me).

• http://sourceforge.net/p/deng/bugs/_discuss/thread/b8af9a2d/b784/attachment/savegame.zip

The underlying issue (that of a degenerate subsector) has been fixed for Monday's build451.

Reopening this issue as this failure case has returned some time after build 451.

I currently suspect this is caused by a bool to boolean conversion somewhere in the BSP node builder as this problem only resurfaced when the boolean type definition changed.

Checked it out in a debugger: a particle with a NULL sector is being processed in linkGeneratorParticles().

There is now an assert() to catch this in Generators_LinkToList().

I think the best we can hope to achieve with this issue at present is to quietly kill any particle which ends up in a degenerate BSP leaf (i.e., sector == NULL as per the assert() in Generators_LinkToList() ).

This particular instance can be resolved but there are many others. The Hexen IWAD is (unfortunately) so riddled with mapping bugs that solving all the degenerate cases is a very tall order indeed.

Pushed a fix that omits all particles in a NULL sector from rendering.