| Summary: | ARB FP shaders appear to have locale-dependant syntax | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Neil Roberts <nroberts> |
| Component: | Mesa core | Assignee: | mesa-dev |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | medium | ||
| Version: | unspecified | ||
| Hardware: | Other | ||
| OS: | All | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Attachments: |
Test program
patch based on Kristian's strtod_l() suggestion Patch to make the lexer use _mesa_strtod Use a strtod() wrapper as we did in the old GLSL compiler |
||
|
Description
Neil Roberts
2009-10-14 11:45:26 UTC
Created attachment 30414 [details] [review] patch based on Kristian's strtod_l() suggestion Can you test this patch? Created attachment 30419 [details] [review] Patch to make the lexer use _mesa_strtod (In reply to comment #1) > Can you test this patch? Thanks. I justed it quickly with the xlib driver (I'm assuming it uses the same code path) but it doesn't work. I think the offending strtod is in src/mesa/shader/program_lexer.l. This directly calls strtod instead of the _mesa wrapper. This patch changes the lexer to use the _mesa wrapper which fixes it for me. It seems a shame to fix it just for glibc sources. There is a float parser in ./src/mesa/drivers/dri/common/xmlconfig.c which isn't affected by the locale, I wonder if we could move that somewhere more accesible and use that instead? I've committed the strtod() -> _mesa_strtod() change to the 7.6 branch. I guess I'd still like to get another opinion on my first patch which uses strtod_l(). Does anyone have any issues with it? (In reply to comment #3) > I've committed the strtod() -> _mesa_strtod() change to the 7.6 branch. > > I guess I'd still like to get another opinion on my first patch which uses > strtod_l(). Does anyone have any issues with it? It looks exactly like that patch that I was intending to write. It looks good to me. OK, committed to 7.6 branch 89b31c9619449d5c9b8ebe4e245c2a926e3583e6 It seems the issue is back with the new compiler. Following is the behaviour with nouveau gallium driver:
~ $ LC_ALL=C ./a.out
strtod("0.5") = 0.500000
nv50_screen_get_param:162 - Unknown PIPE_CAP 11
before glProgramString glGetError() = 0
after glProgramString glGetError() = 0
XIO: fatal IO error 11 (Resource temporarily unavailable) on X server ":0"
after 33 requests (31 known processed) with 0 events remaining.
~ $ LC_ALL=fr_FR.UTF-8 ./a.out
strtod("0.5") = 0,000000
nv50_screen_get_param:162 - Unknown PIPE_CAP 11
before glProgramString glGetError() = 0
after glProgramString glGetError() = 0
XIO: fatal IO error 11 (Ressource temporairement non disponible) on X server ":0"
after 33 requests (31 known processed) with 0 events remaining.
$ glxinfo | grep -i opengl
nv50_screen_get_param:162 - Unknown PIPE_CAP 11
OpenGL vendor string: nouveau
OpenGL renderer string: Gallium 0.4 on NV96
OpenGL version string: 2.1 Mesa 7.9
OpenGL shading language version string: 1.20
OpenGL extensions:
Created attachment 41062 [details] [review] Use a strtod() wrapper as we did in the old GLSL compiler Can you test this patch? (In reply to comment #7) > Created an attachment (id=41062) [details] > Use a strtod() wrapper as we did in the old GLSL compiler > > Can you test this patch? Yes, this fixes the problems I encounter (KDE's blur effect works now). OK, I've committed the patch. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.