Bugzilla

OK, so I'm finally having some luck with reproducing the hang from bug 1598924 (which got this backed out) locally. My initial hypothesis was an asyncshutdown hang related to the blocklist (or something else that got tripped by additional timing), but after doing extensive logging, it seems that is Not It. Instead, having now gotten some locally reproducing things going on and just breaking into a debugger, it seems we're stuck on [Quota Manager's ActorsParent's `SpinEventLoopUntil`](https://searchfox.org/mozilla-central/rev/13b081a62d3f3e3e3120f95564529257b0bf451c/dom/quota/ActorsParent.cpp#3204).

This is sounding vaguely familiar from bug 1594521 where we hit something similar - but I can't tell off-hand if it's really the same issue. This also seems to only affect Windows so pernosco is out, I think? :asuth, any idea how I gather some more debug data here, and/or why we'd be stuck in this event loop for roughly 30 seconds at a time? Happy to add logging or look at specific things when stopping... it's a bit hit-and-miss to reproduce but at the moment I'm having enough luck that hopefully rebuilding and retrying won't be much of a problem.

I'll also note that when this happens, talos as-is ends up killing the launcher process, which leaves the "real" process behind, which then breaks the next start, which is unhelpful, and could probably be the subject of a separate bug.

OK, so I'm finally having some luck with reproducing the hang from bug 1598924 (which got this backed out) locally. My initial hypothesis was an asyncshutdown hang related to the blocklist (or something else that got tripped by additional timing), but after doing extensive logging, it seems that is Not It. Instead, having now gotten some locally reproducing things going on and just breaking into a debugger, it seems we're stuck on [Quota Manager's ActorsParent's `SpinEventLoopUntil`](https://searchfox.org/mozilla-central/rev/13b081a62d3f3e3e3120f95564529257b0bf451c/dom/quota/ActorsParent.cpp#3204).

This is sounding vaguely familiar from bug 1594521 where we hit something similar - but I can't tell off-hand if it's really the same issue. This also seems to only affect Windows so pernosco is out, I think? :asuth, any idea how I gather some more debug data here, and/or why we'd be stuck in this event loop for roughly 30 seconds at a time? Happy to add logging or look at specific things when stopping... it's a bit hit-and-miss to reproduce but at the moment I'm having enough luck that hopefully rebuilding and retrying won't be much of a problem.

Edit: If there is existing NSPR/MOZ_LOG logging that'd be helpful for this that I should turn on, I'd also be happy to hear about it.

I'll also note that when this happens, talos as-is ends up killing the launcher process, which leaves the "real" process behind, which then breaks the next start, which is unhelpful, and could probably be the subject of a separate bug.