nV News Forums


nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   GLX bug? (1.0-3123) (http://www.nvnews.net/vbulletin/showthread.php?t=1870)

marc 09-15-02 02:10 PM

GLX bug? (1.0-3123)
Hi, I'm running an application which tries to compile a (very) long display list (appr 160000 triangles+normals)
with GLX version 2960 this ran fine (and on SGI IR3,V6 and V12 graphics as well), but the exact same code produces the following error with the new GLX (3123) version

gdb reports:
Program received signal SIGSEGV, Segmentation fault.
0x407212c4 in __nvsym14770 () from /usr/lib/libGLcore.so.1

- If I run the application without GL_COMPILE it works fine (but slower...)
- If I generate a smaller surface (a few thousand triangles) there is also no problem.

Is there an (accidental or deliberate) limit to the size of a display list with the new drivers?

(If a friendly NVidia engineer is reading this:
I'm willing to debug the __nvsym14770 call, if you send me the source
;) )

Andy Mecham 09-15-02 02:37 PM

Please send this information, and anything else you think pertinent, to linux-bugs@nvidia.com.



marc 09-21-02 04:58 PM

1 Attachment(s)
I've isolated the offending code fragment
and plugged it in the old boilerplate.c that comes with the mesa sources, and the behaviour is the same:

After 32k triangles (with normals) the 1.0-3123 GLcore library cannot compile the geometry in a display list.

The same executable runs ok on 1.0-2960 (didn't even recompile it)

Do I send this sample code to linux-bugs as well?
(I haven't heard anything on my initial report)

All times are GMT -5. The time now is 03:51 AM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2015, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.