Roman Zavarnitsyn

@looploooop thanks for the report, we'll try to investigate the issue. Out of curiosity, are you actually using performance instrumentation with Sentry or just happened to have this flag enabled...

Thanks a lot for trying the new functionality out, really appreciate that. Unfortunately, at the moment we only support the standard Room setup (i.e. when it's backed by [androidx.sqlite](https://github.com/androidx/androidx/tree/androidx-main/sqlite) lib)....

> Interesting, I thought the bytecode instrumentation was being done at the open helper level, not at the sqlite level itself. Will keep an eye out if this enhancement ever...

What we came up with in the SDK evolution meeting is actually setting a `source` for a span. So we need to investigate whether it's possible to get the database...

This requires a bigger project (we need to have access to *all* `.class` files of the APK, this is not possible with the new AGP API) - potentially we need...

Should be done for 3.0 GA probably? just to make sure, even if it's breaking :)

thanks @gpeal for the report (I just transferred the issue under sentry-gradle-plugin repo as I think it's a better place for it). Indeed, it looks like the task is always...

> check what's the input for the R8 task inside AGP source code, and set the same for this task. Or can we depend on the outputs of R8 task?

So looks like #313 is actually a prerequisite for this one (or rather, implementing #313 would actually fix this issue, so I guess I'd proceed with #313 and then test...

Most likely it would, because we'd use AGP apis to transform the manifest, and they know when to trigger the manifest merging. See recipe here -> https://github.com/android/gradle-recipes/blob/c23620879ef024529605d2b2f7ab27ebadbe55c2/Kotlin/manifestTransformerTest/app/build.gradle.kts#L72-L76 But yeah, it...