The newly released demo of the game Shadowgrounds gives this error: Mesa 7.7-devel implementation error: bad srcFormat in extract float data Please report at bugzilla.freedesktop.org The game menu isn't drawn properly, with black parts hiding a lot of the options. It's possible to start the game, but with the i965 driver the game segfaults (backtrace seems to indicate a problem with the game?) and with the software rasterizer the game is almost completely black (except for the HUD). System environment: -- chipset: G45 / ICH10R -- system architecture: 32-bit -- Linux distribution: Debian unstable -- Machine or mobo model: Asus P5Q-EM -- Display connector: DVI -- KMS: enabled -- xf86-video-intel: efbcf29dd1a1ca058b7a2a93f0685102c06c9369 -- xserver: 1.6.3.901 -- mesa: 5ce72c559c027ea03513ed9c0038706733b0586e -- drm: 67e4172394a88d4922fb8d9c7c3d96ce7e02c5a6 -- kernel: 2.6.31-rc9
Can you set a breakpoint at line 3715 of extract_float_rgba (in src/mesa/main/image.c). I'd be interested to see the value of srcFormat and a backtrace.
Created attachment 29674 [details] Backtrace from break Sure, I hope this is correct: Breakpoint 1, extract_float_rgba (n=512, rgba=0xbffee83c, srcFormat=34682, srcType=5121, src=0xb5f99008, swapBytes=0 '\0') at main/image.c:3717 3717 main/image.c: No such file or directory. in main/image.c
Created attachment 29676 [details] [review] Call _mesa_unpack_dudv_span_byte with the base internal format I think this patch should fix the issue that you encountered. Brian and Roland, Could one or both of you review this patch please?
The "implementation error" message is gone now, but the problems with the game are the same.
(In reply to comment #3) > Created an attachment (id=29676) [details] > Call _mesa_unpack_dudv_span_byte with the base internal format > > I think this patch should fix the issue that you encountered. > > Brian and Roland, > > Could one or both of you review this patch please? This looks good to me. I'm not sure what the game is doing makes a whole lot of sense though as it's submitting unsigned byte data. I guess if it's only needing positive offsets that's ok (but wouldn't make any sense with shaders). (In reply to comment #4) > The "implementation error" message is gone now, but the problems with the game > are the same. If that's a problem related to that texture format, you could try setting the env var MESA_EXTENSION_OVERRIDE=-GL_ATI_envmap_bumpmap in this case the game should never try to use that texture format.
(In reply to comment #5) > If that's a problem related to that texture format, you could try setting the > env var MESA_EXTENSION_OVERRIDE=-GL_ATI_envmap_bumpmap in this case the game > should never try to use that texture format. Thanks, I tried this but it makes no difference.
Created attachment 29719 [details] backtrace from crash with i965 Forgot to attach this earlier, this is the backtrace from the crash which happens with the i965 driver.
Created attachment 29720 [details] screenshot when software rendering is forced
The rendering errors in the menu are still there (and the warning about an implementation error) but the actual game no longer crashes and seems to be playable. I'm using git master, a3abb9d1cc4698e45e5d794919e8c27a3d7e5908
Can you update your git tree and run again? I just committed a patch that'll give more info in the "bad srcFormat" warning.
Sure, the warning is now: Mesa 7.8-devel implementation error: bad srcFormat GL_DU8DV8_ATI in extract float data
Created attachment 32441 [details] [review] proposed patch to fix the warning/problem report Can you try this patch? It looks like the GL_ATI_envmap_bumpmap spec allows GL_DU8DV8_ATI as a <format> parameter to glTexImage2D. It should probably only be allowed as an <internalFormat> parameter. Oh well.
The warning is gone now, thanks! The problems with the menu are still there though, so I guess that's an unrelated bug.
I've committed the patch to the 7.7 branch. I don't know what the problem is with the menu.
Created attachment 32445 [details] screenshot of the menu faded to black The menu problem seems to have somthing to do with S3TC? With libtxc_dxtn.so the menu will slowly fade almost completely black, without it, it fades to white. I have attached a screenshot of the menu, black and unreadable.
I can work around the fade-to-black problem by always setting haveAlpha = GL_TRUE in encodedxtcolorblockfaster() in libtxc_dxtn, I have no idea how to fix this properly, but I thought it worth mentioning in case it can help somebody track it down.
Hey cool! Marek's fixes in http://cgit.freedesktop.org/~cbrill/libtxc_dxtn/commit/?id=32772908bc9b759a6b285e77af96419b4c9c8feb solves this problem. Do I dare hope for a new release of libtxc_dxtn?
(In reply to comment #17) > Hey cool! > > Marek's fixes in > http://cgit.freedesktop.org/~cbrill/libtxc_dxtn/commit/?id=32772908bc9b759a6b285e77af96419b4c9c8feb > solves this problem. Do I dare hope for a new release of libtxc_dxtn? I announced new libtxc_dxtn on mesa-dev. Closing..
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.