Status Update
Comments
br...@snapchat.com <br...@snapchat.com> #2
Thanks for the report. I will route this to the appropriate internal team and update this when I hear back from them.
mk...@google.com <mk...@google.com> #3
br...@snapchat.com <br...@snapchat.com> #4
"2022-06-12 18:47:15.156 1841-4562/? W/PackageManager: Intent does not match component's intent filter: Intent { act=com.google.android.gms.wearable.BIND_LISTENER"
mk...@google.com <mk...@google.com> #5
ap...@google.com <ap...@google.com> #6
+1, can confirm it doesn't work on Android 13:=
2022-07-15 11:26:15.023 589-5347 PackageManager pid-589 W Intent does not match component's intent filter: Intent { act=com.google.android.gms.wearable.BIND_LISTENER cmp=xxx/xxx.WatchMessageReceiver }
2022-07-15 11:26:15.023 589-5347 PackageManager pid-589 W Access blocked: ComponentInfo{xxx/xxx.WatchMessageReceiver}
2022-07-15 11:26:15.023 589-5347 ActivityManager pid-589 W Unable to start service Intent { act=com.google.android.gms.wearable.BIND_LISTENER cmp=xxx/xxx.WatchMessageReceiver } U=0: not found
ap...@google.com <ap...@google.com> #7
Note that I've been able to make it work by:
- Adding
<action android:name="com.google.android.gms.wearable.BIND_LISTENER" />
in the intent filter - Removing
<data android:scheme="wear" android:host="*" />
But I feel like this is not something we should do
br...@snapchat.com <br...@snapchat.com> #8
I'm really afraid Android 13 might get released as-is, breaking WearOS app communication 😨😨
ap...@google.com <ap...@google.com> #9
If you're not targeting API 33 you're not affected by the bug. So it's a big bug, and yes we of course expected more from Google, but you can always target the api level later when it's fixed.
But I agree this is kind of desperating that more than 1.5 month after the first report nothing has changed.
ap...@google.com <ap...@google.com> #10
As an interim update on this issue: we've been already working on the fix that should be available by Android 13 release. The fix requires thorough testing, I'll keep this bug updated as soon as we have more to share. Thanks!
ap...@google.com <ap...@google.com> #11
@
Thank you for the update @
ap...@google.com <ap...@google.com> #12
Android 13 is out today and we still have no patch unlike what you said a month ago
ap...@google.com <ap...@google.com> #13
ap...@google.com <ap...@google.com> #14
This issues has been already given high priority (updated external priority on this bug to reflect internal status). The fix is on the way and going through the final rounds of testing, so the roll out is slated to next couple of weeks.
To reiterate what have been mentioned earlier on this bug: this issue affects only apps targeting Android 13, so the apps won't break unless you bump targetSDK
version to 33
. In case if you want to start working on app compatibility for Android 13 behaviour changes, you could use
mk...@google.com <mk...@google.com> #15
- The report is 2 months old
- Google chose to release Android 13 with that bug
- There's no mention of this bug on the documentation so you can totally bump your targetSdk without noticing it
So thank you guys for working on this but it's still not a valid excuse for taking that long for such an important issue. Now that being said, let us know when a fix is available
br...@snapchat.com <br...@snapchat.com> #16
mk...@google.com <mk...@google.com> #17
That must be some really intense testing as we are 10 days later and still nothing on sight. I don't want to be a P2 issue if that's what a P1 is.
Description
Background
We upgraded from AGP 7.2.2 to AGP 7.4.2 and are now seeing hard class verification failures (crash on startup) on Android 11. It appears the root cause is due to the ApiModelOutliner now being enabled and having a weird interaction with our existing ApiHelper classes (as described here ).
I'll share artifacts (dex2oat logs, compiler dump) over email.
Details
Original Code:
Proguard Rules:
Smali Before (AGP 7.2):
Smali After (AGP 7.4):
Workaround
We can fix the crash by removing our proguard rules; however, we see that it will cause class verification failures to be re-introduced in other spots. In other words, ApiModelOutliner does not yet appear to be 100% effective so we'd need to keep our approach for the time being and thus revert back to AGP 7.2 in the interim.