New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 616088 link

Starred by 2 users

Issue metadata

Status: Duplicate
Owner: ----
Closed: Aug 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 3
Type: Bug



Sign in to add a comment

jp.naver.line.android crashing a lot in a way that doesn't decode

Project Member Reported by torne@chromium.org, May 31 2016

Issue description

See https://crash.corp.google.com/browse?q=product.name%3D%27AndroidWebView%27%20AND%20special_protos.user_feedback.mobile_data.build_data.sdk_int%3E%3D21%20AND%20special_protos.user_feedback.mobile_data.package_data.package_name%3D%27jp.naver.line.android.100500.8590045411%27%20AND%20product.version%3D%2750.0.2661.86%27&ignore_case=false&enable_rewrite=true&omit_field_name=&omit_field_value=&omit_field_opt=%3D#samplereports:5,cpuarchitecture for all crashes in the app.

See also https://crash.corp.google.com/browse?q=product.name%3D%27AndroidWebView%27%20AND%20special_protos.user_feedback.mobile_data.build_data.sdk_int%3E%3D21%20AND%20special_protos.user_feedback.mobile_data.package_data.package_name%3D%27jp.naver.line.android.100500.8590045411%27%20AND%20product.version%3D%2750.0.2661.86%27%20AND%20custom_data.ChromeCrashProto.magic_signature_1.name%3D%27abort_message%27&ignore_case=false&enable_rewrite=true&omit_field_name=&omit_field_value=&omit_field_opt=%3D&stbtiq=&reportid=&index=0 for the most common crash in the app, which has a bad stack (stops at std::terminate and doesn't unwind any further).

There's two problems here:

1) Why doesn't this stack unwind properly? It's only ARM where it ends up like this; on x86 there's a stack that's presumably the same issue and unwinds further (into std::bad_alloc exception, which is probably what's actually happening on arm too). All addresses here were found by stack scanning, so it should have been able to just keep on scanning and find more.

2) Why does this app crash so much with OOM errors? Test team, is this reproducible? If so then we may be able to use this to investigate the general OOM issues we're running into.
 

Comment 2 by torne@chromium.org, Jun 2 2016

Yes, that's expected - we don't ever have the symbols for system libraries.

Comment 3 by hush@chromium.org, Jun 3 2016

Cc: satyavat...@chromium.org
+ Satya
> 2) Why does this app crash so much with OOM errors? Test team, is this reproducible? If so then we may be able to use this to investigate the general OOM issues we're running into.

Comment 4 by boliu@chromium.org, Jun 13 2016

we have pretty good repro for OOM now (more or less):  crbug.com/619173 

Is there anything else to do here?

Comment 5 by torne@chromium.org, Jun 14 2016

We should work out why this stack doesn't unwind. There's nothing special about this crash or this app as far as I can tell, and I would expect a proper stack unwinding that the magic signature processor can tell is an OOM.

Comment 6 by torne@chromium.org, Jun 14 2016

(also, just one repro for the OOM isn't necessarily enough, as it may happen in many different ways).
Labels: -Needs-Bisect
Tried to repro this issue and also tried using the steps in the  crbug.com/619173  on Zperia Z3/Android 6.0/23.5.A.0.575 and Z3+(with multiples apps opened). No luck.

Note issue 633313 might be related here.

Comment 9 by ti...@chromium.org, Aug 8 2016

Cc: torne@chromium.org
Shall we close this as non-actionable? I do not see much info in the reports...
Cc: boliu@chromium.org
leaving it to torne, but I guess should just dup into crbug.com/633313, unless there is interest in improving breakpad for c++ exceptions? probably not very useful?
Mergedinto: 633313
Status: Duplicate (was: Unconfirmed)
Yeah now we figured out that it's actually an uncaught exception it's less interesting. If we see more exceptions then generally the thing to do is kill code that throws exceptions until they stop :p

Sign in to add a comment