Made logging macros evaluate its stream arguments lazily.

Made logging macros evaluate its stream arguments lazily.

Also added logging unittests.

BUG=57383
TEST=logging_unittest.cc

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=61127

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=61142

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=61249

[akalin, 2010-09-30 05:47:43]

+evmar for review

[Evan Martin, 2010-09-30 18:46:54]

This LG as best as I can read it

http://codereview.chromium.org/3575008/diff/7002/13004
File base/logging.h (left):

http://codereview.chromium.org/3575008/diff/7002/13004#oldcode500
base/logging.h:500: 
This diff is nearly impossible to read since code is moved as well as changed. 
:(

In the future, please do code moves as one change and then code edits as a
followup.

http://codereview.chromium.org/3575008/diff/7002/13004
File base/logging.h (right):

http://codereview.chromium.org/3575008/diff/7002/13004#newcode269
base/logging.h:269: // LOG_DFATAL_LEVEL is LOG_FATAL in debug mode, ERROR in
normal mode
s/_LEVEL//  ?

http://codereview.chromium.org/3575008/diff/7002/13004#newcode328
base/logging.h:328: // Helper macro which avoids evaulating the arguments to a
stream if
typo: evaluating

http://codereview.chromium.org/3575008/diff/7002/13004#newcode444
base/logging.h:444: logging::LogMessage(__FILE__, __LINE__, _result).stream()
Curiously, code like

if (foo)
  CHECK(bar);
else
  baz;

Will actually do the wrong thing (the else will be bound to the ifs inside the
macro expansion).  But I guess this just has never come up!

[Evan Martin, 2010-09-30 18:50:07]

This LG as best as I can read it

[akalin, 2010-09-30 19:27:54]

Please take another look (with side-by-side diff)!

http://codereview.chromium.org/3575008/diff/7002/13004
File base/logging.h (left):

http://codereview.chromium.org/3575008/diff/7002/13004#oldcode500
base/logging.h:500: 
On 2010/09/30 18:46:54, Evan Martin wrote:
> This diff is nearly impossible to read since code is moved as well as changed.

> :(
> 
> In the future, please do code moves as one change and then code edits as a
> followup.

Sorry it's hard to read. :( It's hard to avoid crappy diffs when doing pretty
big changes like these.

I found that the side-by-side diff is easier to follow.  Mind taking another
look with that?

http://codereview.chromium.org/3575008/diff/7002/13004
File base/logging.h (right):

http://codereview.chromium.org/3575008/diff/7002/13004#newcode269
base/logging.h:269: // LOG_DFATAL_LEVEL is LOG_FATAL in debug mode, ERROR in
normal mode
On 2010/09/30 18:46:54, Evan Martin wrote:
> s/_LEVEL//  ?

Done.

http://codereview.chromium.org/3575008/diff/7002/13004#newcode328
base/logging.h:328: // Helper macro which avoids evaulating the arguments to a
stream if
On 2010/09/30 18:46:54, Evan Martin wrote:
> typo: evaluating

Done.

http://codereview.chromium.org/3575008/diff/7002/13004#newcode444
base/logging.h:444: logging::LogMessage(__FILE__, __LINE__, _result).stream()
On 2010/09/30 18:46:54, Evan Martin wrote:
> Curiously, code like
> 
> if (foo)
>   CHECK(bar);
> else
>   baz;
> 
> Will actually do the wrong thing (the else will be bound to the ifs inside the
> macro expansion).  But I guess this just has never come up!

CHECK() would work fine, but CHECK_EQ() wouldn't.  This should actually be easy
to fix, just by changing to:

if (!LOG_IS_ON(...)) {} else ...

I'll do that in a follow-up CL.

[Evan Martin, 2010-09-30 22:15:11]

LG stands

[akalin, 2010-10-01 00:00:28]

Fixed a few issues revealed by build breakage, resubmitting once trybots pass.

On 2010/09/30 22:15:11, Evan Martin wrote:
> LG stands