I'm dogfooding gfx.webrender.software with a Win32 Firefox build (on Win64 OS). I don't see the tab crash reporter, but I get periodic info bars popping up for "unsent crash reports". This suggests that background tabs are crashing (2-4 at a time).

I see multiple crash signatures that point to SW-WR code:

bp-c7b431d4-a640-4b46-82a5-875e90200920 [@ Composite]
bp-1c5ed48e-aa0f-44c3-b061-2fc960200920 [@ core::ops::function::Fn::call<T> | trunc]
bp-b9eb5222-663a-42ee-93db-e8bbf0200920 [@ memcpy | scale_blit]

Top 10 frames of crashing thread:

0 xul.dll Composite gfx/wr/swgl/src/composite.h:426
1 xul.dll webrender_bindings::swgl_bindings::SwCompositeThread::process_job gfx/webrender_bindings/src/swgl_bindings.rs:652
2 xul.dll std::sys_common::backtrace::__rust_begin_short_backtrace<closure-0,  ../4fb7144ed159f94491249e86d5bbd033b5d60550/src/libstd/sys_common/backtrace.rs:130
3 xul.dll core::ops::function::FnOnce::call_once<closure-0,  ../4fb7144ed159f94491249e86d5bbd033b5d60550/src/libcore/ops/function.rs:232
4 xul.dll alloc::boxed::{{impl}}::call_once< ../4fb7144ed159f94491249e86d5bbd033b5d60550/src/liballoc/boxed.rs:1017
5 xul.dll std::sys::windows::thread::{{impl}}::new::thread_start ../4fb7144ed159f94491249e86d5bbd033b5d60550//src/libstd/sys/windows/thread.rs:51
6 kernel32.dll BaseThreadInitThunk 
7 mozglue.dll patched_BaseThreadInitThunk mozglue/dllservices/WindowsDllBlocklist.cpp:592
8 ntdll.dll _RtlUserThreadStart 
9 ntdll.dll _RtlUserThreadStart 

I see this on Windows, it's the GPU process that's crashing, and then you get switched to Basic Layers (and no GPU process) after it's crashed 4 times.

https://hg.mozilla.org/mozilla-central/rev/0ee0283c751a

I'm still seeing this on a build that should have this fix.

This is crashing on try pushes as well.

https://hg.mozilla.org/mozilla-central/rev/b9294db776bf

Matt, I hit this sw-wr crash twice today with Nightly build 20201009041754, but this bug was resolved fixed on 2020-10-02. Might your fix have missed some corner case? Or perhaps this be a new regression?

[@ memcpy | scale_blit ] bp-7ad4fc66-e899-411f-9fd3-d47a60201009
[@ core::ops::function::Fn::call<T> | trunc ] bp-136c1fdc-f55c-470d-ac46-8b9b30201009

The second of those seems unrelated (it's a crash elsewhere in WR), but the first looks like another variation of this bug.

Any idea what the STR might be to reproduce that?

(In reply to Matt Woodrow (:mattwoodrow) from comment #15)

The second of those seems unrelated (it's a crash elsewhere in WR), but the first looks like another variation of this bug.

OK. I'll remove the second signature from this bug. There is an existing bug 1606580 (filed ten months ago) with the second signature, but it looks unrelated because [@ RustMozCrash] is a very generic signature. I might file a new bug for this particular crash.

Any idea what the STR might be to reproduce that?

I don't know. I don't see content crash reporter for these crashes, so I don't know what I was doing when they happened. They're found by the unsubmitted crash report scan when I restart Nightly each morning. That suggests they're background tab or shutdown crashes.

If this crash happens again, I'll file a new bug (instead of reopening this one).