Crash report: https://crash-stats.mozilla.org/report/index/0501f93c-fc91-4da4-9c68-abe0e0210216

MOZ_CRASH Reason: MOZ_RELEASE_ASSERT(false) (Unhandled external image format)

Top 10 frames of crashing thread:

0 xul.dll mozilla::wr::RenderTextureHostSWGL::LockSWGLCompositeSurface gfx/webrender_bindings/RenderTextureHostSWGL.cpp:173
1 xul.dll mozilla::wr::wr_swgl_lock_composite_surface gfx/webrender_bindings/RenderTextureHostSWGL.cpp:191
2 xul.dll webrender_bindings::bindings::{{impl}}::lock_composite_surface gfx/webrender_bindings/src/bindings.rs:1439
3 xul.dll webrender::compositor::sw_compositor::{{impl}}::add_surface gfx/wr/webrender/src/compositor/sw_compositor.rs:1626
4 xul.dll webrender::renderer::Renderer::draw_frame gfx/wr/webrender/src/renderer/mod.rs:4398
5 xul.dll webrender::renderer::Renderer::render_impl gfx/wr/webrender/src/renderer/mod.rs:2154
6 xul.dll webrender::renderer::Renderer::render gfx/wr/webrender/src/renderer/mod.rs:1890
7 xul.dll webrender_bindings::bindings::wr_renderer_render gfx/webrender_bindings/src/bindings.rs:634
8 xul.dll mozilla::wr::RendererOGL::UpdateAndRender gfx/webrender_bindings/RendererOGL.cpp:186
9 xul.dll mozilla::wr::RenderThread::UpdateAndRender gfx/webrender_bindings/RenderThread.cpp:481

It seems that fallback from WebRender to WR-WR seem to be related. Old TextureHost might be used.

(In reply to Sotaro Ikeda [:sotaro] from comment #1)

It seems that fallback from WebRender to WR-WR seem to be related. Old TextureHost might be used.

If that's the case, bug 1688096 will fix this since I unify the SW/HW textures types. There may be more work for texture types I hadn't considered though since I was focused on Windows -- specifically RenderEGLImageTextureHost. Android isn't a problem since we don't allow fallback to SW WR or SW WR popups.

Looks like RenderEGLImageTextureHost is used on Android, not Linux:

https://searchfox.org/mozilla-central/rev/0379f315c75a2875d716b4f5e1a18bf27188f1e6/gfx/gl/SharedSurface.cpp#139

That leaves RenderDMABUFTextureHost as the last one we might need to worry about.

Set release status flags based on info from the regressing bug 1689203

I will fix this, but there shouldn't be a need to uplift. We only fallback from HW-WR to SW-WR in nightly. Starting in SW-WR won't have this problem.

When the crash happened, format seemed to be SurfaceFormat::UNKNOWN: (

This crash signature seems to have disappeared in recent nightlies. I guess bug 1688096 did fix it?

(In reply to Julien Cristau [:jcristau] from comment #8)

This crash signature seems to have disappeared in recent nightlies. I guess bug 1688096 did fix it?

Yep.