Change the behavior of --enable-dcheck from crashing

Change the behavior of --enable-dcheck in release
build from crashing to logging the failure and 
continuing. In addition,

- In interactive mode, we will display a message
 box so the user can either click OK and continue
 or attach a debugger. 
- If Chrome is running in test environment, it
 has the option to overide the default behavior
 and suppress the error dialog.

To support this, a new severity level ERROR_REPORT
 is added.

In addition, DFATAL is now mapped to ERROR_REPORT
instead of ERROR. The only usage of DFATAL is in
ssl_client_socket_nss.cc. Add wtc for code review.



Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=9633

[Nicolas Sylvain, 2009-02-10 18:45:38]

http://codereview.chromium.org/21216/diff/1/2
File base/logging.h (right):

http://codereview.chromium.org/21216/diff/1/2#newcode398
Line 398: LOG_IF(ERROR, !(condition)) << "Check failed: " #condition ". "
Is it still going to display a message box?

[Nicolas Sylvain, 2009-02-11 18:27:57]

Looks like I was crazy. The message box was disabled in revision 4844 by brettw,
so this is not a regression.

LGTM

[huanr, 2009-02-11 23:32:51]

wtc, you only need to review whether the change of DFATAL mapping is reasonable.

[huanr, 2009-02-11 23:34:18]

The change with its description has been updated. Nicolas, can you take another
look?

[Nicolas Sylvain, 2009-02-12 00:15:17]

quick question: Are you going to set a log report handler in your case?

Nicolas

http://codereview.chromium.org/21216/diff/5/1005
File base/logging.cc (left):

http://codereview.chromium.org/21216/diff/5/1005#oldcode455
Line 455: DCHECK(r != WAIT_ABANDONED);
why did you remove this line?

[Nicolas Sylvain, 2009-02-12 00:40:26]

LGTM

[wtc, 2009-02-19 01:32:03]

Huan,

I didn't notice this code review until today.  Sorry.
It looks good to me.  Not sure why I have anything to
say about the DFATAL mapping.

I have some suggested fixes for nits below.  Since logging.h
is included by a lot of files, I'm not sure if it's worth
fixing these nits.

http://codereview.chromium.org/21216/diff/2001/3002
File base/logging.cc (right):

http://codereview.chromium.org/21216/diff/2001/3002#newcode94
Line 94: // An report handler override specified by the client to be called
instead of
Nit: add a blank line above this line.

http://codereview.chromium.org/21216/diff/2001/3001
File base/logging.h (right):

http://codereview.chromium.org/21216/diff/2001/3001#newcode86
Line 86: // Note the special severity of ERROR_REPORT only available/relevant in
normal
Nit: add "is" before "available/relevant".

http://codereview.chromium.org/21216/diff/2001/3001#newcode87
Line 87: // mode, which displays error dialog without terminating the program.
There is
Nit: add "an" before "error dialog".

http://codereview.chromium.org/21216/diff/2001/3001#newcode159
Line 159: // The default handler shows a dialog box and then terminate the
process,
Nit: "terminate" => "terminates"

http://codereview.chromium.org/21216/diff/2001/3001#newcode164
Line 164: // Sets the Log Report Handler that will be used to notify of check
failures
Nit: add a blank line above this line.