Closed Ristovski closed 7 years ago
I think this is the key message:
Error - Couldn't choose default framebuffer config
Which comes from here in the code. After that it might be a knock-on effect causing the other failures and probably the segfault.
I'm not sure what to make of this as I've never seen that fail before. You hardware sounds capable enough so maybe there's an issue somewhere in the software configuration. I'm not requesting anything outlandish like a depth/stencil buffer or anything, it's just an 8-bit double buffered sRGB backbuffer that should be supported anywhere. All the complex rendering happens off-screen.
Can you try modifying that code to take out some of the attribs and see by elimination which one is the problem? Also could you post the output of your glxinfo?
It might also help to use renderdoccmd replay proton.rdc
for testing since then you don't go all the way through the Qt UI, and just do a simple window setup and preview.
This will be interesting :smiley:
glxinfo
- Intel: https://bpaste.net/raw/0aa580f17ec1
glxinfo
- Radeon: https://bpaste.net/raw/4b6405fc3879
Confirmed as mesa bug: https://bugs.freedesktop.org/show_bug.cgi?id=102264
Committed a workaround that falls back to a non-SRGB fbconfig, tested to work.
Edit: This has been identified as a mesa issue, see last comment.
Using renderdoc both from
0.x
and1.x
branches results in a segmentation fault after Renderdoc finishes loading the frames after the application exits.Edit: This only happens with OpenGL, Vulkan works fine.
Before the segfault occurs, renderdoc spits out this rather lengthy log: https://bpaste.net/raw/ab68561dd5e2
The gist of this issue might be related to "Error - Couldn't choose default framebuffer config"
Running via
gdb
yields the following:This issue was also present the first time I tried using renderdoc which was maybe half a year ago.
Happens both when running on Intel integrated and dedicated AMD.
Specs: Intel HD 4400 Haswell "Gen7.5" Radeon HD 8850M "Southern Islands" Xorg/mesa/llvm/clang all from git