Initialize Android CrashDumpManager earlier

Initialize Android CrashDumpManager earlier

The CrashDumpManagerAndroid instance must be initialized before any
child process is created. This change moves the initialization to
PreThreadsCreate, which at the very least should be before the
PROCESS_LAUNCHER thread (where the potentially evil child processes will
be created). #nocrashes

BUG=439322
Committed: https://crrev.com/0a0e4fddc32adad7c90f10dce2eb7c57ef652e89
Cr-Commit-Position: refs/heads/master@{#314083}

[cjhopman, 2015-01-31 01:01:31]

cjhopman@chromium.org changed reviewers:
+ thestig@chromium.org

[cjhopman, 2015-01-31 01:01:32]

thestig: *

[Lei Zhang, 2015-01-31 01:14:22]

https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
File components/crash/browser/crash_dump_manager_android.cc (right):

https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
components/crash/browser/crash_dump_manager_android.cc:39: CHECK(!instance_);
I haven't looked very hard, but are you worried this code isn't getting
exercised enough as is because it's only executed in official builds without
DCHECKs?

[cjhopman, 2015-01-31 01:20:48]

https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
File components/crash/browser/crash_dump_manager_android.cc (right):

https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
components/crash/browser/crash_dump_manager_android.cc:39: CHECK(!instance_);
On 2015/01/31 01:14:22, Lei Zhang wrote:
> I haven't looked very hard, but are you worried this code isn't getting
> exercised enough as is because it's only executed in official builds without
> DCHECKs?

Thinking about it a bit, I think this one could probably just stay a DCHECK, I'm
not really worried about us calling this twice.

I think I'd keep the two added CHECKs though so that if we have these crashes
again sometime, it'll be obvious that it's because this hasn't been initialized.

[Lei Zhang, 2015-01-31 01:49:15]

On 2015/01/31 01:20:48, cjhopman wrote:
>
https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
> File components/crash/browser/crash_dump_manager_android.cc (right):
> 
>
https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
> components/crash/browser/crash_dump_manager_android.cc:39: CHECK(!instance_);
> On 2015/01/31 01:14:22, Lei Zhang wrote:
> > I haven't looked very hard, but are you worried this code isn't getting
> > exercised enough as is because it's only executed in official builds without
> > DCHECKs?
> 
> Thinking about it a bit, I think this one could probably just stay a DCHECK,
I'm
> not really worried about us calling this twice.
> 
> I think I'd keep the two added CHECKs though so that if we have these crashes
> again sometime, it'll be obvious that it's because this hasn't been
initialized.

But can we even get info on the crash, if we crash in the CrashDumpManager? X_X

[cjhopman, 2015-01-31 01:57:52]

On 2015/01/31 01:49:15, Lei Zhang wrote:
> On 2015/01/31 01:20:48, cjhopman wrote:
> >
>
https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
> > File components/crash/browser/crash_dump_manager_android.cc (right):
> > 
> >
>
https://codereview.chromium.org/889083002/diff/1/components/crash/browser/cra...
> > components/crash/browser/crash_dump_manager_android.cc:39:
CHECK(!instance_);
> > On 2015/01/31 01:14:22, Lei Zhang wrote:
> > > I haven't looked very hard, but are you worried this code isn't getting
> > > exercised enough as is because it's only executed in official builds
without
> > > DCHECKs?
> > 
> > Thinking about it a bit, I think this one could probably just stay a DCHECK,
> I'm
> > not really worried about us calling this twice.
> > 
> > I think I'd keep the two added CHECKs though so that if we have these
crashes
> > again sometime, it'll be obvious that it's because this hasn't been
> initialized.
> 
> But can we even get info on the crash, if we crash in the CrashDumpManager?
X_X

Yes, we can (go/crash/fb59a65a2e05e9be for example). Really, the
CrashDumpManagerAndroid just handles creating a file for child processes to
write their minidump to and then some minimal processing of those minidumps to
prepare them for uploading. It's not involved in reporting browser crashes. 

Maybe it should be renamed to reflect that...

[Lei Zhang, 2015-01-31 02:09:57]

Got it, LGTM

[cjhopman, 2015-01-31 02:12:11]

The CQ bit was checked by cjhopman@chromium.org

[commit-bot: I haz the power, 2015-01-31 02:12:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/889083002/20001

[commit-bot: I haz the power, 2015-01-31 08:42:57]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-01-31 08:42:57]

Try jobs failed on following builders:
  win8_chromium_gn_rel on tryserver.chromium.win
(http://build.chromium.org/p/tryserver.chromium.win/builders/win8_chromium_gn_...)

[Lei Zhang, 2015-02-01 04:09:20]

The CQ bit was checked by thestig@chromium.org

[commit-bot: I haz the power, 2015-02-01 04:09:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/889083002/20001

[commit-bot: I haz the power, 2015-02-01 05:41:56]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-02-01 05:42:55]

Patchset 2 (id:??) landed as
https://crrev.com/0a0e4fddc32adad7c90f10dce2eb7c57ef652e89
Cr-Commit-Position: refs/heads/master@{#314083}