Breakpad Linux: Log crashes.

Breakpad Linux: Log crashes.

BUG=41106
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=75303

[Lei Zhang, 2011-02-15 03:59:52]

Stuart: The time stamp is in milliseconds. Is that right?

[agl, 2011-02-15 14:33:52]

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc
File chrome/app/breakpad_linux.cc (right):

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc...
chrome/app/breakpad_linux.cc:659: FilePath tmp_path("/tmp");
We certainly can't write files like this into /tmp. How about the user data dir?

[stuartmorgan, 2011-02-15 15:45:14]

On 2011/02/15 03:59:52, Lei Zhang wrote:
> Stuart: The time stamp is in milliseconds. Is that right?

It's seconds on the Mac, actually (whole seconds), so it would be nice to be
consistent.

[Lei Zhang, 2011-02-15 20:29:48]

Changed timestamps to seconds.

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc
File chrome/app/breakpad_linux.cc (right):

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc...
chrome/app/breakpad_linux.cc:659: FilePath tmp_path("/tmp");
On 2011/02/15 14:33:52, agl wrote:
> We certainly can't write files like this into /tmp. How about the user data
dir?

DIR_CRASH_DUMPS is relative to DIR_USER_DATA. If a user cannot get
DIR_CRASH_DUMPS, then they probably also cannot get DIR_USER_DATA. In which
case, Chrome will refuse to start.

[agl, 2011-02-15 20:34:14]

LGTM

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc
File chrome/app/breakpad_linux.cc (right):

http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc...
chrome/app/breakpad_linux.cc:659: FilePath tmp_path("/tmp");
On 2011/02/15 20:29:48, Lei Zhang wrote:
> DIR_CRASH_DUMPS is relative to DIR_USER_DATA. If a user cannot get
> DIR_CRASH_DUMPS, then they probably also cannot get DIR_USER_DATA. In which
> case, Chrome will refuse to start.

In the case that we can't get DIR_CRASH_DUMPS, I think that we should log and
return. Writing into /tmp is open to a symlink attack and the current code could
become a security issue in the event of a distant change.

[Lei Zhang, 2011-02-15 21:08:25]

On 2011/02/15 20:34:14, agl wrote:
> LGTM
> 
> http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc
> File chrome/app/breakpad_linux.cc (right):
> 
>
http://codereview.chromium.org/6526010/diff/1002/chrome/app/breakpad_linux.cc...
> chrome/app/breakpad_linux.cc:659: FilePath tmp_path("/tmp");
> On 2011/02/15 20:29:48, Lei Zhang wrote:
> > DIR_CRASH_DUMPS is relative to DIR_USER_DATA. If a user cannot get
> > DIR_CRASH_DUMPS, then they probably also cannot get DIR_USER_DATA. In which
> > case, Chrome will refuse to start.
> 
> In the case that we can't get DIR_CRASH_DUMPS, I think that we should log and
> return. Writing into /tmp is open to a symlink attack and the current code
could
> become a security issue in the event of a distant change.

Sure. If we can't get DIR_CRASH_DUMPS, then |crash_log_path| will remain NULL
and we won't write to the log.

[Lei Zhang, 2011-02-16 03:13:09]

Stuart: if you can just take one more look. I'll commit this tomorrow if I can.

[stuartmorgan, 2011-02-17 19:27:58]

LGTM