Re: [Dri-devel] Neverwinter Nights.

[Roland S. <rsc...@sw...>]

Adam K Kirchhoff wrote:
> A few weeks ago there was some discussion about the fact that the
> post-texmem merge code has broken Neverwinter Nights.
I believe you are speaking about radeon/r200?
The post-texmem merge did not break Neverwinter Nights - it just made 
some errors which were there before more apparent (i.e. the texture 
flickering increased, the lighting is still wrong in exactly the same 
way as before).
The two issues (texture flickering/color flashes and wrong lighting) 
seem to be completely independant (in fact, I completely eliminated the 
former by just decreasing the R200_CMD_BUF_SIZE to 1024, unfortunately I 
don't have a clue why that even works (possibly only works for me)).
Unfortunately it looks like nobody is actively investigating these 
issues, probably none of the developers have that game (though there are 
issues in glaze3d which could be related to the former problem) & time 
to investigate - I've tried for the first, but I'm lost. And without 
docs I can't even start to figure out the second.
But anyway, if you want to play nwn you can still use R200_NO_TCL=1 (or 
the equivalent of that for the radeon) which makes all problems 
disappear. Hurts framerate a lot, unfortunately (at least with my "slow" 
Athlon XP1600/sdr).
Screenshots can still be found here:
(correct, without tcl)
http://homepage.hispeed.ch/rscheidegger/nwn_without_tcl.jpg
(with tcl)
http://homepage.hispeed.ch/rscheidegger/nwn_with_tcl.jpg
(with 1k buffer hack)
http://homepage.hispeed.ch/rscheidegger/nwn_tcl_1kbuf.jpg

> Has the exact cause
> been determined?  If so, what needs to be done to get this working again?
Or are you speaking of the problem that nwn can't get a visual at 
startup and just craps out? Leif Delgass has posted a patch for that 
(bug in reporting visuals) but it was decided it needs more testing as 
it could possibly trigger some unwanted side effects (works fine here 
though). (This problem was indeed introduced by the texmem-merge, as 
previously the driver just didn't fail when it couldn't match the 
requested buffersize (which is against the spec but possibly there just 
to avoid the problem with the wrongly reported visuals).)

http://sourceforge.net/mailarchive/forum.php?thread_id=2054806&forum_id=7177 
(I think you are aware of that already however, as you've participated 
in that thread ;-)).
If you don't want to apply the patch, you could of course start the game 
via gdb, set a breakpoint at the code when it searches for a glx visual 
and change one variable when it hits the breakpoint ;-)

Roland