Bug 40212: Disable more telemetry (especially, since it caused crashes)
This is another patch for #40212 (closed), that is intended also to replace my brutal patch.
It solves the crash on my test device and on my daily driver (both of them experienced them).
I think that the cause of the crash was Glean not initialized/only partially initialized, during a call to Glean.setUploadEnabled(true)
(see app/src/main/java/org/mozilla/fenix/components/metrics/GleanMetricsService.kt
).
I'm not 100% satisfied by this patch, because I encountered a huge amount of these lines in logcat:
05-18 12:09:49.904 2624 2660 I glean/Dispatchers: Task queued for execution and delayed until flushed
at a certain point they are replaced by these ones:
05-18 12:09:49.904 2624 2660 E glean/Dispatchers: Exceeded maximum queue size, discarding task
which means that they're not being sent (good).
Maybe good enough for the alpha? I propose that we merge, tag and start build4 from it (since we have the problem with the addons and mozconfig-android-all
anyway).
In Glean's Rust code I've seen that some parts use an unbounded channel, but from the above error I'd say we are not risking to fill the memory.
Anyway, I think that we should investigate a little bit more, but I fear that fixing these errors require modifications to glean, which we currently take directly as a third-party package.
While writing the text for this MR, I've noticed that these Glean logs are also on the debug build. This means we might be able to get to them.