Open DerArtem opened 2 years ago
From gdb I am only getting this backtrace:
Thread 1 "cog" received signal SIGSEGV, Segmentation fault.
0xb251f5e8 in enable_tile_status () from target:/usr/lib/libEGL.so.1
(gdb) bt
#0 0xb251f5e8 in enable_tile_status () from target:/usr/lib/libEGL.so.1
#1 0xb263cebc in ffi_call_SYSV () from target:/usr/lib/libffi.so.8
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
(gdb)
@DerArtem I suppose this is the board, which according to this PDF comes with a Vivante GPU, and I assume the screen output is configured to use the built-in MIPI DSI interface, in which case would use imx-drm
. Can you confirm this is indeed your setup?
I suspect this configuration has stopped working after https://github.com/Igalia/cog/pull/371 or (less likely) https://github.com/Igalia/cog/pull/392—these have been the biggest changes in the DRM plug-in in the last year.
OTOH, the GLES renderer should work without glitching in this configuration, so I think that grants a separate issue report.
@DerArtem Can you try disabling atomic mode setting? Something like this should do:
cat > cog.conf <<EOF
[drm]
disable-atomic-modesetting=false
EOF
cog -C cog.conf -P drm $URL
If the above causes glitches, then we know that the GLES renderer is glitching because currently it does not currently use atomic modesetting. If that works and doesn't glitch, then we know we have some issue in the atomic modesetting handling in the modeset
renderer.
@DerArtem One more question: are you using Etnaviv or the Vivante proprietary GPU driver?
Hello,
on my STM32MP1 i am getting a Segmentation fault when running cog. This seems to occour when something on the display changes. Eg, when I open a simple html website it is displayed on the screen. As soon as I click on something it crashes. Using gles renderer the issue does not occur, but I have graphic glitches.
On cog 0.12.4 this problem does not exist.
Here is a log of the segfault: