[Mac] Log some NSException info with crash dumps.

[Mac] Log some NSException info with crash dumps.

Adds two breakpad keys.  "firstexception" will store the name and
reason for the first NSException caught by the event loop.
"lastexception" will store the same for the last one seen before
crashing.

http://crbug.com/24462
TEST=Hard to test without known exceptions to handle.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=29349

[John Grabowski, 2009-10-14 18:16:52]

LGTM

Please also follow-up using ideas from Neal; e.g.
BreakpadGenerateAndSendReport() 

Perhaps add DCHECK in here for developer builds to be more... illustrative?

http://codereview.chromium.org/275008/diff/1/2
File chrome/browser/chrome_application_mac.mm (right):

http://codereview.chromium.org/275008/diff/1/2#newcode156
Line 156: NSString* value = [NSString stringWithFormat:@"%@ reason %@",
Can we find a way to do this that doesn't allocate memory?

[Scott Hess - ex-Googler, 2009-10-16 17:42:37]

jrg also wrote:
> Please also follow-up using ideas from Neal; e.g.
> BreakpadGenerateAndSendReport() 
> 
> Perhaps add DCHECK in here for developer builds to be more... illustrative?

Am working on those.  Deferring the DCHECK because it doesn't have a push
deadline associated with it.  Still experimenting with forcing a report.

http://codereview.chromium.org/275008/diff/1/2
File chrome/browser/chrome_application_mac.mm (right):

http://codereview.chromium.org/275008/diff/1/2#newcode156
Line 156: NSString* value = [NSString stringWithFormat:@"%@ reason %@",
On 2009/10/14 18:16:52, John Grabowski wrote:
> Can we find a way to do this that doesn't allocate memory?

I think we should be conservative and push what we have.  If we find that it's
excessive, we can always back it down to just -reason or something, but getting
a week of backtraces with firstexception == "NSGenericException" would be
frustrating.