Glean iOS emits "duplicate ping named 'events' or 'baseline' etc. as errors
Categories
(Data Platform and Tools :: Glean: SDK, defect, P1)
Tracking
(Not tracked)
People
(Reporter: travis_, Assigned: travis_)
References
Details
Glean appears to be emitting a lot of "duplicate" pings being registered errors on iOS. This is showing up in the Firefox iOS Sentry logs as noise. Hopefully there is some way to avoid the error altogether, or at least suppress it as a warning or info.
|
||
Comment 1•5 years ago
|
||
We should do a proper audit of the log levels. Errors should only be actionable errors, everything else should probably be lower. And that "duplicate" line might just be a debug log
|
|
||
Updated•5 years ago
|
|
Assignee | |
Comment 2•5 years ago
|
||
Hmm, it looks like we already log these at debug level. Which makes me think that we either need to change this to info, or perhaps the real problem is that we do indeed register the internal pings twice. We manually register the internal pings in the Swift bindings synchronously in the initialize function. We also initialize these pings asynchronously through the dispatcher when the Ping constructor is called.
[:janerik] What do you think about checking in the Ping.init code for the internal ping names that are synchronously registered in initialize and just skipping the registration in the Ping constructor for these pings?
|
||
Comment 3•5 years ago
|
||
(In reply to Travis Long [:travis_] from comment #2)
Hmm, it looks like we already log these at
debuglevel. Which makes me think that we either need to change this to info, or perhaps the real problem is that we do indeed register the internal pings twice.
Is sentry supposed to report 'debug' logs?! I mean, aside from the possibility of a double-registration bug on our end, should it really be capturing all the debug logs as problems in Sentry? Maybe this is really a Sentry configuration issue, as we can't possibly think of removing all the logs, right? :D
|
|
||
Comment 4•5 years ago
|
||
You know what? We dropped it from error to debug level just last month. We did do a release, but that one might not have made it to consumers.
(In reply to Travis Long [:travis_] from comment #2)
[:janerik] What do you think about checking in the
Ping.initcode for the internal ping names that are synchronously registered in initialize and just skipping the registration in the Ping constructor for these pings?
That's bascially what this check does anyway. Registering twice is harmless.
Closing this as it's been done already.
Description
•