TTimo / GtkRadiant

The open source, cross platform level editor for idtech games
http://icculus.org/gtkradiant/
Other
588 stars 155 forks source link

RTCW SP Shader Overflow #223

Open 0bsidian opened 10 years ago

0bsidian commented 10 years ago

Several RTCW single player mappers are saying that they still rely on Q3Map instead of Q3Map2 because of a bug that occurs on larger maps. Sounds like it only occurs with Q3Map2 and not with the original Q3Map.

It has been documented here: http://www.katsbits.com/tutorials/idtech/q3map2-error-messages.php?subaction=showfull&id=1148502027&archive=&start_from=&ucat=15&

-SSF-Sage has provided a sample map that he says produces this error: http://zero.robotrenegade.com/radiant/tobruk_shaderoverflow.pk3

SzicoVII commented 10 years ago

This isnt unique to RTCW - I've had this many times in JKA in MP and SP - using complex lighting, low lightmapscales, -samples 3 in compiles combined with a detailled map causes an almost identical problem, except it isn't just weapon flashes, you see random images everywhere, sometimes it looks like a screenshot of your player from 3rd person view, others looks like a void e.t.c.

A fix would be AMAZING, its been plaguing the lighting quality in my maps for years.

It seems to be linked to the length of time the compiler spends on the last stage of the light compile - "allocating" step of StoreSurfaceLightmaps. Apparently this stage isn't multi-threaded, which it could probably do with being.

ensiform commented 10 years ago

Its most likely dual-ly broken with q3map2 and the engines themselves. Also, JA uses screen and scratch buffers and stuff for glow within a game image pointer so thats why you might see the player image.

SzicoVII commented 10 years ago

Possibly...but Ive found that using a better PC with more RAM and a modified q3map2 to utilise more allows you to go further before experiencing the error which suggests to me its an issue with the way q3map2 stores the lightmaps in the ram before baking them into the bsp.