Bugzilla

(In reply to Henrik Skupin [:whimboo][⌚️UTC+2] from comment #8)
> This caused massive test failures recently for automation because Firefox in most cases didn't shutdown. While this is fixed for harnesses like Marionette and Remote Protocol now, what about real users?

Unless the worker here is explicitly being spun up by something in the shutdown process, users are unlikely to encounter the problem, and they definitely should not experience it.

> Would they experience the same hang during shutdown? If yes, should this maybe block the 117 release?

As I Understand It (AIUI), the tests are turning orange because the test infrastructure is correctly detecting a child content process shutdown hang, but this is only something that's possible for debug/similar builds; in production/non-debug builds we don't care about whether content processes shutdown cleanly and don't bother even trying to shut them down fully.

(In reply to Henrik Skupin [:whimboo][⌚️UTC+2] from comment #8)
> This caused massive test failures recently for automation because Firefox in most cases didn't shutdown. While this is fixed for harnesses like Marionette and Remote Protocol now, what about real users?

Unless the worker here is explicitly being spun up by something in the shutdown process, users are unlikely to reproduce the problem (translation worker racing shutdown), and they definitely should not experience it (a hang).

> Would they experience the same hang during shutdown? If yes, should this maybe block the 117 release?

As I Understand It (AIUI), the tests are turning orange because the test infrastructure is correctly detecting a child content process shutdown hang, but this is only something that's possible for debug/similar builds; in production/non-debug builds we don't care about whether content processes shutdown cleanly and don't bother even trying to shut them down fully.

(In reply to Henrik Skupin [:whimboo][⌚️UTC+2] from comment #8)
> This caused massive test failures recently for automation because Firefox in most cases didn't shutdown. While this is fixed for harnesses like Marionette and Remote Protocol now, what about real users?

Unless the worker here is explicitly being spun up by something in the shutdown process, users are unlikely to reproduce the problem (translation worker racing shutdown), and they definitely should not experience the implications of any hangs, as what I see reported is only for content processes.  Also, see the next para.

> Would they experience the same hang during shutdown? If yes, should this maybe block the 117 release?

As I Understand It (AIUI), the tests are turning orange because the test infrastructure is correctly detecting a child content process shutdown hang, but this is only something that's possible for debug/similar builds; in production/non-debug builds we don't care about whether content processes shutdown cleanly and don't bother even trying to shut them down fully.