WebRender device resets and broken rendering
Categories
(Core :: Widget: Gtk, defect)
Tracking
()
People
(Reporter: heftig, Unassigned)
References
(Regression)
Details
(Keywords: regression)
Attachments
(2 files)
Sometimes while browsing, suddenly Firefox spews a bunch of messages to stdout and falls back to software WebRender. Rendering of the browser window is then often also broken until it is resized.
The log messages vary, but usually contain lines like these:
1:48.60 INFO: b'[GFX1-]: Detect DeviceReset DeviceResetReason::UNKNOWN DeviceResetDetectPlace::WR_POST_UPDATE in Parent process'
1:49.33 INFO: b'[GFX1-]: Failed to compile vertex shader: ps_text_run_ALPHA_PASS_TEXTURE_2D'
1:49.33 INFO: b'[ERROR webrender::device::gl] Failed to compile vertex shader: ps_text_run_ALPHA_PASS_TEXTURE_2D'
... more failed shader compilations ...
1:49.34 INFO: b'[GFX1-]: wr_renderer_render: Shader(Compilation("composite_FAST_PATH_TEXTURE_2D", ""))'
1:49.34 INFO: b'[GFX1-]: Handling webrender error 2'
1:49.34 INFO: b'[GFX1-]: Fallback WR to SW-WR'
1:49.36 INFO: b'[GFX1-]: Detect DeviceReset DeviceResetReason::UNKNOWN DeviceResetDetectPlace::WR_POST_UPDATE in Parent process'
I found a reliable way to reproduce this:
- Start Firefox
- Wait about 10 seconds for the cursor in the address bar to stop blinking
- Wait 1 minute longer
- Press Ctrl+N to open a new window
The log spew happens. The old window then corrupts when attempting to scroll. The new window is sometimes also broken.
I don't remember this happening months ago, but the regression search ended like this:
Last good revision: cd13ab11acd1d8720f9824ae5d2f6bfcfbdf8f5c
First bad revision: 1c5835148de52e3821bc5ad92ef159342fcc1cc8
Pushlog: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=cd13ab11acd1d8720f9824ae5d2f6bfcfbdf8f5c&tochange=1c5835148de52e3821bc5ad92ef159342fcc1cc8
Firefox Nightly
Arch Linux
Mesa 25.2.3
Linux 6.16.7
AMD Radeon 780M Graphics (radeonsi, phoenix, LLVM 20.1.8, DRM 3.64, 6.16.7-zen1-1-zen)
Comment 1•7 months ago
|
||
Set release status flags based on info from the regressing bug 1959645
:stransky, since you are the author of the regressor, bug 1959645, could you take a look?
For more information, please visit BugBot documentation.
| Reporter | ||
Comment 2•7 months ago
|
||
| Reporter | ||
Comment 3•7 months ago
|
||
| Reporter | ||
Comment 4•7 months ago
|
||
Just confirmed that release 143.0 is affected as well.
| Reporter | ||
Comment 5•7 months ago
|
||
This was also filed on the AMD DRM bugtracker at https://gitlab.freedesktop.org/drm/amd/-/issues/4568.
Updated•7 months ago
|
Comment 7•7 months ago
|
||
I wonder if that's Firefox bug which was triggered by MESA change or MESA bug revealed by Firefox. We also have NVIDIA proprietary driver bug caused by Bug 1959645.
| Reporter | ||
Comment 8•7 months ago
|
||
Since the other bug also reproduces this with 142, I was mistaken and the regression I found only works for these exact reproduction steps. Still, maybe something can be learned from this?
Updated•6 months ago
|
Description
•