Looks like runining a local mochites is broken right now, it fails with:

1:03.43 CRASH: pid:30933. Test:RemoteProcessMonitor. Minidump analysed:False. Signature:[@ nsJARChannel::CreateJarInput(nsIZipReaderCache*, nsJARInputThunk**)]
Mozilla crash reason: Missing chrome or resource URLs: resource://gre/modules/CrashService.jsm

robwu found this.

CrashService.jsm is registered in components.conf: https://searchfox.org/mozilla-central/rev/5a362eb7d054740dc9d7c82c79a2efbc5f3e4776/toolkit/components/crashes/components.conf#11

CrashService.jsm and components.conf are registered in moz.build, conditional on MOZ_CRASHREPORTER: https://searchfox.org/mozilla-central/rev/5a362eb7d054740dc9d7c82c79a2efbc5f3e4776/toolkit/components/crashes/moz.build#18,25

I think that what happened is:

• Artifact build objects are built with MOZ_CRASHREPORTER set (--enable-crashreporter), so the component registration is included in the compiled objects.
• Local build doesn't have ac_add_options --enable-crashreporter, so the CrashService.jsm is not bundled.

Result: The local artifact build is unusable

The check that triggers the crash was introduced in 1721627

(In reply to Rob Wu [:robwu] from comment #1)

I think that what happened is:

• Artifact build objects are built with MOZ_CRASHREPORTER set (--enable-crashreporter), so the component registration is included in the compiled objects.
• Local build doesn't have ac_add_options --enable-crashreporter, so the CrashService.jsm is not bundled.

Result: The local artifact build is unusable

Why doesn't the same thing happen with desktop, and can we fix either the local or artifact builds for geckoview here so the configuration is consistent? Nick, do you know?

(In reply to :Gijs (low availability; he/him) from comment #3)

(In reply to Rob Wu [:robwu] from comment #1)

I think that what happened is:

• Artifact build objects are built with MOZ_CRASHREPORTER set (--enable-crashreporter), so the component registration is included in the compiled objects.
• Local build doesn't have ac_add_options --enable-crashreporter, so the CrashService.jsm is not bundled.

Result: The local artifact build is unusable

Why doesn't the same thing happen with desktop, and can we fix either the local or artifact builds for geckoview here so the configuration is consistent? Nick, do you know?

Historically, we couldn't cross-compile breakpad from particular hosts (I think it was macOS) to Android targets: see https://searchfox.org/mozilla-central/rev/ad2ffab089e4e0c0fe99a1a046ab2b1c45546bdb/old-configure.in#1740 and the tickets linked from there. I don't know if that's still the case; certainly nobody will have tried for A Good Long Time.

To address this, we can either remove the cross-compiling restriction entirely (and do some testing to see what happens) or we can enable the crash reporter for Android artifact builds across the board, baking in this detail of the CI configuration.

It would be nicer if the former Just Worked, but the latter is certainly an option.

People keep hitting this, resetting priority to discuss during triage.

(In reply to Agi Sferro | :agi | ni? for questions | ⏰ PST | he/him from comment #6)

People keep hitting this, resetting priority to discuss during triage.

I think that we should work-around this particular issue to unblock things, as I suggested in https://bugzilla.mozilla.org/show_bug.cgi?id=1728226#c4. But I also just filed Bug 1751762, which is a proposal to rework some fundamental assumptions around artifact builds that would pave a path to solving this type of issue in a systematic manner.

We don't actually support building artifacts with no crashreporter so we should
set a sensible default.

Set release status flags based on info from the regressing bug 1721627

https://hg.mozilla.org/mozilla-central/rev/4cf7d64258f1