595167 - Make Breakpad client libs work on Android

Status: RESOLVED FIXED

(Toolkit :: Crash Reporting, defect)

Description

[(not currently active) Ted Mielczarek]

The first step in getting crash reporting working on Android will be making sure the Breakpad libs build and work on Android.

Comment 1

[(not currently active) Ted Mielczarek]

Progress report:
I've got the Breakpad client code compiling and linking for Android, and I had Vivien run it for me on a Nexus One (it fails to run on my Tegra for some reason). It's failing a bunch of unit tests, but that's not surprising. There's a bit more work to make all the unit tests pass, I'd estimate maybe another week once I get an N1 to test on.

Comment 2

[(not currently active) Ted Mielczarek]

I fixed some assumptions in the unit tests and most of them pass now (thanks to Vivien for running it again). There are just a few failing tests which I'll be unable to diagnose until I get a device that can run the tests, but it's probably working well enough now that I can try building fennec with it and seeing what happens.

Current patch is here:
http://hg.mozilla.org/users/tmielczarek_mozilla.com/breakpad-mq/file/511553dc4d22/android

Comment 3

[(not currently active) Ted Mielczarek]

I built Fennec with this patch, and was able to get a minidump out using "Crash me now!" There are still some issues, but that's very promising!

Comment 4

[(not currently active) Ted Mielczarek]

Attachment: Make Breakpad client libs work on Android

Comment 5

[(not currently active) Ted Mielczarek]

Actually some bits of this patch ostensibly belong on bug 595171. I just mashed them up so I could test on device, since the Breakpad unittests don't run on my Tegra.

Comment 6

[(not currently active) Ted Mielczarek]

This is working, but File ID calculation is failing for some reason. It works in the unit tests, whether I run them as a straight C++ program, or loaded into an Android app using JNI, but fails when run inside of Fennec. That needs to be fixed since otherwise we can't match modules up to symbol files.

Comment 7

[(not currently active) Ted Mielczarek]

Ah, I fixed it. I think this is working well enough to work on getting patches upstream.

Comment 8

[(not currently active) Ted Mielczarek]

The longer explanation is that instead of taking a full Breakpad update snapshot, I just ported my patch on top of the existing Breakpad code in mozilla-central. Unfortunately the file ID code there exhibited a bug that the upstream code didn't, since the upstream code had since been changed.

Comment 9

[(not currently active) Ted Mielczarek]

Upstream patch up for review:
http://breakpad.appspot.com/212001/show

Comment 10

[(not currently active) Ted Mielczarek]

linux_syscall_support.h got split out into its own repository, since it's shared with Chromium as well, and I had to patch it too:
http://codereview.chromium.org/3656004

Comment 11

[(not currently active) Ted Mielczarek]

Attachment: Make Breakpad client libs work on Android

Comment 12

[(not currently active) Ted Mielczarek]

Comment on attachment 482864 [details] [diff] [review]
Make Breakpad client libs work on Android

These are just the Breakpad changes, merged to our local Breakpad snapshot (the upstream patch doesn't apply cleanly).

Comment 13

[(not currently active) Ted Mielczarek]

This landed upstream:
http://code.google.com/p/google-breakpad/source/detail?r=716

I'll leave this bug open until we land it in m-c.

Comment 14

[Brad Lassey [:blassey] (use needinfo?)]

did you mean to clear the blocking flag?

Comment 15

[(not currently active) Ted Mielczarek]

No, that was probably just bugzilla screwing me.

Comment 16

[(not currently active) Ted Mielczarek]

Attachment: Make Breakpad client libs work on Android.

Comment 17

[Brad Lassey [:blassey] (use needinfo?)]

pushed http://hg.mozilla.org/mozilla-central/rev/ca8226c47fea
with two follow ups for bustage fixes:
http://hg.mozilla.org/mozilla-central/rev/df5fd0d9d783 and
http://hg.mozilla.org/mozilla-central/rev/7b324b328cbc

Comment 18

[(not currently active) Ted Mielczarek]

What was failing that you had to check those fixes in? An Android build, or a desktop Linux build? (I'll need to make sure those wind up upstream.)

Comment 19

[Brad Lassey [:blassey] (use needinfo?)]

(In reply to comment #18)
> What was failing that you had to check those fixes in? An Android build, or a
> desktop Linux build? (I'll need to make sure those wind up upstream.)

> http://hg.mozilla.org/mozilla-central/rev/df5fd0d9d783
fixes http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287621662.1287625046.2209.gz which effected all linux builders
> http://hg.mozilla.org/mozilla-central/rev/7b324b328cbc
fixes http://tinderbox.mozilla.org/showlog.cgi?log=Mobile/1287625133.1287625394.3995.gz which effected half of the maemo builders (filed bug 606179 for the inconsistency)

Comment 20

[(not currently active) Ted Mielczarek]

(In reply to comment #17)
> http://hg.mozilla.org/mozilla-central/rev/7b324b328cbc

I upstreamed a fix for this (not exactly the same patch, but it fixes it):
http://code.google.com/p/google-breakpad/source/detail?r=718

Comment 21

[(not currently active) Ted Mielczarek]

(In reply to comment #17)
> pushed http://hg.mozilla.org/mozilla-central/rev/ca8226c47fea
> with two follow ups for bustage fixes:
> http://hg.mozilla.org/mozilla-central/rev/df5fd0d9d783 and

Posted this one for upstream review here:
http://codereview.chromium.org/4053002