Created attachment 89577 [details]
Screenshot of Dota 2 after changing resolutions attached. The menu "layers" occasionally flicker into view with no apparent (to my eyes) timing, seemingly randomly.
Changing the resolution back to the original does not fix it. Quitting the game and restarting it starts the game properly in the resolution changed into, and everything works properly.
Mesa version is 9.2.2 from Debian Jessie, i386-arch package as Dota 2 is a 32-bit application. The system otherwise is amd64. Gnome3 as window manager. GPU is Intel Iris Pro.
I've just assumed this is a game bug. Have you taken an apitrace and replayed it on another driver?
Happens also on Ubuntu with latest (couple of week old) Mesa sources.
The issue happens after applying any video option changes, not just resolution (advanced setting changes aren't applied until DotA2 is restarted, this issue happens after changing resolution, vsync etc generic video options).
DOTA2 outputs a long list of extensions that are not supported with Mesa when it starts. If I would need to guess which of them missing and DOTA2 not handling that correctly could be related to this, maybe GL_NV_fence or GLX_EXT_swap_control_tear.
AMD doesn't support NV_fence either, and GLX_swap_control_tear shouldn't be necessary.
I second Eric's comment: it'd be great to try this on non-Mesa drivers.
Does this occur on Mesa 10.0 or git?
(In reply to comment #4)
> Does this occur on Mesa 10.0 or git?
It happens also with yesterday's git master version of Mesa.
I'll send you apitrace of the issue privately (it's few hundred MBs). With Mesa Intel driver trace flickers, with Nvidia proprietary driver it would seem to work fine (is there some way to slow down trace replay on faster cards?).
Apitrace on Nvidia complains about memory waste: texture 15 has mipmaps, while its min filter is inconsistent with mipmaps.
With i965 Mesa driver it complains about invalid operation in glBindFrameBuffer(buffer), and a bit about GPU stalls on mapping busy object.
This bug is also exist on Mesa10.4RC1
This issue is still exist on mesa-10.4.0.
(In reply to fangxun from comment #7)
> This issue is still exist on mesa-10.4.0.
I was not able to reproduce this bug with mesa-10.3.2. But in my case I have a haswell GPU.
Petri, could you mark this as fixed if you cannot reproduce it anymore with recent Mesa / Dota version?
(In reply to Eero Tamminen from comment #9)
> Petri, could you mark this as fixed if you cannot reproduce it anymore with
> recent Mesa / Dota version?
Let's do this the opposite way. NEEDINFO until the reporter can confirm and reopen.
This Mesa bug has been in the "NEEDINFO" status for over 60 days. I am closing this bug based on lack of response but feel free to reopen if resolution is still needed. Please ensure you're supplying the correct information as requested.