Re: Reasons processes might fail to attach to ActivityManagerService

[Ubuntu guy]

How reproducible is that? Any memory errors in klogs?

On Apr 2, 5:19 am, Kyle Russell <krussell7...@gmail.com> wrote:
> What are some reasons that a process might fail to attach itself to the
> ActivityManagerService?  I can see where ActivityManagerService starts the
> process, but a little later I get notified that it fails to attach.  I'm
> working with 4.2.2-r1.
>
> I/ActivityManager( 2329): Start proc android.process.acore for content
> provider com.android.providers.contacts/.ContactsProvider2: pid=2472
> uid=10000 gids={50000, 3003, 1015, 1028}
> ...
> W/ActivityManager( 2329): Process ProcessRecord{411b6328
> 2472:android.process.acore/u0a10000} failed to attach
>
> I don't see any obvious logcat messages in between the start proc attempt
> and failure that would explain why it didn't (couldn't?) attach.  Once the
> failure happens, ActivityManagerServices tries to restart the process in
> question, but this seems to be a cycle it never overcomes in order to
> finish boot.

[Ubuntu guy]

Is this a bring up effort on a new hardware? If so, whats mem config?
I have seen issues while trying to port Honeycomb on a 512 MB ram in
the past but yes it wasn't as bad as kills during boot. You may also
want to refer and make sure of memory requirements as specified in
CDD (http://static.googleusercontent.com/external_content/
untrusted_dlcp/source.android.com/en/us/compatibility/4.2/android-4.2-
cdd.pdf)

On Apr 3, 10:02 am, Kyle Russell <krussell7...@gmail.com> wrote:
> It's very reproducible.  I don't see any memory-related errors before the
> "failed to attach" messages in logcat or dmesg, but eventually (after
> ActivityManagerService tries a few restarts) the OOM driver starts killing
> processes.
>
> It only seems to be related to a few processes
> too: com.android.launcher, com.android.settings, android.process.acore, com.android.inputmethod.latin, com.android.location.fused,
> plus a couple of others.  Granted that may have more to do with the fact
> that it happens early in boot.
>
> The first process where I see a "failed to attach" message isn't always the
> same either.  Sometimes I never get past the boot animation; sometimes I
> get all the way to the launcher.  However, when the launcher does start, it
> only stays up until OOM kills it, so something else bad is going on in the
> background.

[Kyle Russell]

[Kyle Russell]

It's very reproducible.  I don't see any memory-related errors before the "failed to attach" messages in logcat or dmesg, but eventually (after ActivityManagerService tries a few restarts) the OOM driver starts killing processes.

It only seems to be related to a few processes too: com.android.launcher, com.android.settings, android.process.acore, com.android.inputmethod.latin, com.android.location.fused, plus a couple of others.  Granted that may have more to do with the fact that it happens early in boot.

The first process where I see a "failed to attach" message isn't always the same either.  Sometimes I never get past the boot animation; sometimes I get all the way to the launcher.  However, when the launcher does start, it only stays up until OOM kills it, so something else bad is going on in the background.

On Tuesday, April 2, 2013 7:27:44 PM UTC-4, Ubuntu guy wrote: