On Apr 1, 2014, at 8:46 AM, Mark Finkle <[email protected]> wrote:
> One of our highest ranking crashes is a "java.lang.UnsatisfiedLinkError"
> crash. It mainly shows on the Google Play Developer console. We have a
> variety of bugs [1] [2] [3] [4] filed about it, but it still is a bit of a
> mystery.
>
> Recently, we started to think that the problem is caused by the wrong APK
> trying to run on the wrong ABI (ARM vs x86). We've had a few crash reporters
> work with us to try different APKs and the problem clears up. One reporter
> switched from the Google Play delivered APK to an explicit sideload of the
> expected-correct APK and the crash went away.
If we can somehow prove this is what happens (preferably by reproducing it on
an actual device), we should go to Google directly. Pretty unacceptable bug,
IMHO. Do we have one of those Xperia things to try it there?
I have a hard time believing this happens to other companies. I wonder if there
is something in our Play setup that is causing this behavior.
>
> If we continue with this hypothesis, we have a few questions/actions:
> 1. How are the wrong APKs being delivered to devices by Google Play? I have
> not noticed any problems or bugs related to Google Play about this issue.
> Could this also be happening on 3rd party stores?
If it was installed via third party, we wouldn’t see the crash reports in Play,
right?
> 2. We should start explicitly checking CPU before loading the native
> libraries. We can inform the user of the problem and even collect Telemetry.
Definitely.
> 3. Figure out a way to fix the issue in-place, either by (a) directing the
> user to the correct APK or (b) downloading the correct libraries.
>
> #3a might not be effective since we have no way of forcing an APK download
> from Google Play and pointing people to a Mozilla hosted APK means that
> sideloading needs to be enabled on the device. Maybe sideloading is not the
> end of the world though.
Both of these are pretty terrible, honestly. People will wonder why their
Firefox is updating differently from other stuff, etc.
James
_______________________________________________
mobile-firefox-dev mailing list
[email protected]
https://mail.mozilla.org/listinfo/mobile-firefox-dev
