Attempt at fixing crash during importing. As far as I can tell from...

Attempt at fixing crash during importing. As far as I can tell from
the crash the BookmarkModel is trying to notifying observers it's
going away and one of those observers has been deleted. This is during
first run import, so the only observer should be the importer.

This may be possible if someone prematurely exits the nested loop
started during first run so that the importerhost isn't done when the
profile is deleted.

BUG=16143
TEST=none


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=28227

[cpu_(ooo_6.6-7.5), 2009-10-02 17:32:11]

.

[cpu_(ooo_6.6-7.5), 2009-10-02 17:33:48]

We've seen crashes like this in the past. I've always wonder who could be
quitting our nested message loop.


On 2009/10/02 17:32:11, cpu wrote:
> .

[cpu_(ooo_6.6-7.5), 2009-10-02 17:42:58]

Ignore my comment. I need coffee.

[cpu_(ooo_6.6-7.5), 2009-10-05 23:09:28]

rahul, he indeed is trying to solve the crashing at first run, when the import
is done in the subprocess.

I just have one question below

http://codereview.chromium.org/242119/diff/1/2
File chrome/browser/importer/importer.cc (right):

http://codereview.chromium.org/242119/diff/1/2#newcode542
Line 542: profile_ = target_profile;
do we ever reset profile_ to NULL, maybe when we finish importing?

[sky, 2009-10-05 23:13:12]

http://codereview.chromium.org/242119/diff/1/2
File chrome/browser/importer/importer.cc (right):

http://codereview.chromium.org/242119/diff/1/2#newcode542
Line 542: profile_ = target_profile;
On 2009/10/05 23:09:28, cpu wrote:
> do we ever reset profile_ to NULL, maybe when we finish importing?
> 

No, once set we never null it out.
I was tempted to plumb the constructor to take the Profile, but that proves
tricky in a couple of cases.

[cpu_(ooo_6.6-7.5), 2009-10-06 02:38:50]

LGTM

Since rahul is the main importer guy I'll deffer to him for any objection.

[kuchhal, 2009-10-06 17:26:27]

lgtm based on offline conversation.

The bug http://code.google.com/p/chromium/issues/detail?id=16143 points to
strange crash which doesn't look like is in importer. The crash server seems to
point to this line (brower_main.cc:453):

return ResultCodes::NORMAL_EXIT_EXP1;