Created attachment 109567 [details] Screen shot of the issue, from a trimmed trace It seems that one of the dota2 updates over the last month or so introduced a regression causing visual glitches as shown in the attached files. Originally I've assumed to be a mesa regression, but it seems that's not the case - old, previously working mesa produces the same output. MESA_EXTENSION_OVERRIDE=-GL_ARB_draw_elements_base_vertex makes no difference, plus replaying a trace recorded on i965/snb (no issues there) exhibits the same glitches. Note: After trimming the trace some textures are not missing and the text appears block. The trace plus a screenshot of a non-trimmed trace will follow shortly.
Created attachment 109568 [details] Screen shot of the issue, from original trace Reference image of the issue from a non-trimmed trace.
The trace is here: http://people.freedesktop.org/~evelikov/fdo86356/dota2.trace.xz
The following messages while replaying the trace are due to the trimming: Mesa: User error: GL_INVALID_OPERATION in glDeleteShader XX: glDebugOutputCallback: High severity API error 2, GL_INVALID_OPERATION in glDeleteShader 0 XX glDeleteShader(shader = YY) Running against debug mesa (./configure --enable-debug ...) does not trigger any asserts or produce extra output to stderr/stdout.
mogorva on IRC confirms that this bug is the same as the TXL-related one. Even though it's newer, it has all the relevant info, so marking as dup. *** This bug has been marked as a duplicate of bug 90887 ***
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.