Bug 1843782 Comment 58 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

I've tried to get egl device selection to resolve this problem (by picking a mesa software device) but I have not gotten it to work despite extensive experimentation and debugging, it always ends up failing to set up the actual rendering context with that device and hence webgl gets disabled (which is admittedly a safe outcome but not one we want).

The other method we have within easy reach is to set the environment variable LIBGL_ALWAYS_SOFTWARE=1, but we need to set that before any threads are created (because setenv is not very thread-safe), which I am not sure is the case when we detect this hardware, detection may need to happen earlier in startup if threads are going at this point.  We could just try comment #43 and hope that setenv doesn't cause a crash.  XUL may have a thread-safe wrapper for setenv/getenv however and I am going to be looking for that now (I think I saw one) that I have pivoted to this approach.

I am actively working on a patch for this that I do want to land in firefox 120 (and of course esr115) before the disclosure deadline.
I've tried to get egl device selection to resolve this problem (by picking a mesa software device) but I have not gotten it to work despite extensive experimentation and debugging, it always ends up failing in eglChooseConfig to find any configs and hence webgl gets disabled (which is admittedly a safe outcome but not one we want).

The other method we have within easy reach is to set the environment variable LIBGL_ALWAYS_SOFTWARE=1, but we need to set that before any threads are created (because setenv is not very thread-safe), which I am not sure is the case when we detect this hardware, detection may need to happen earlier in startup if threads are going at this point.  We could just try comment #43 and hope that setenv doesn't cause a crash.  XUL may have a thread-safe wrapper for setenv/getenv however and I am going to be looking for that now (I think I saw one) that I have pivoted to this approach.

I am actively working on a patch for this that I do want to land in firefox 120 (and of course esr115) before the disclosure deadline.

Back to Bug 1843782 Comment 58