Fixed numerous issues that were causing errors in profiler log processing

Fixed numerous issues that were causing errors in profiler log processing
on a real web application loaded in the test shell.

Also implemented output of JSON-encoded call stacks for profiler
prototype.

Committed: http://code.google.com/p/v8/source/detail?r=1646

[Søren Thygesen Gjesse, 2009-03-31 07:23:13]

http://codereview.chromium.org/56064/diff/1/2
File src/log.cc (right):

http://codereview.chromium.org/56064/diff/1/2#newcode405
Line 405: OS::FSetMode(Logger::logfile_, false);
I think we should try to avoid the file mode change (see comments in
platform.h).

http://codereview.chromium.org/56064/diff/1/2#newcode435
Line 435: msg.WriteCStringToLogFile(content);
Will this still be required if we use binary file mode?

http://codereview.chromium.org/56064/diff/1/6
File src/platform.h (right):

http://codereview.chromium.org/56064/diff/1/6#newcode157
Line 157: static void FSetMode(FILE* stream, bool text_mode);
I think we should try to avoid having this function. How about making sure that
on windows "b" is added to the mode when performing fopen on Windows. That
should avoid any translation when writing to files on Windows. Having just \n on
Windows is preferable to adding this function and trying to control the file
mode while writing.

http://codereview.chromium.org/56064/diff/1/7
File src/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/56064/diff/1/7#newcode828
Line 828: LOG(RegExpCodeCreateEvent("RegExp", *code, *source));
Is the first parameter required for this RegExp specific function?

[Mikhail Naganov, 2009-03-31 08:52:51]

Thanks for insightful comments!

http://codereview.chromium.org/56064/diff/1/2
File src/log.cc (right):

http://codereview.chromium.org/56064/diff/1/2#newcode405
Line 405: OS::FSetMode(Logger::logfile_, false);
On 2009/03/31 07:23:13, Søren Gjesse wrote:
> I think we should try to avoid the file mode change (see comments in
> platform.h).

Done.

http://codereview.chromium.org/56064/diff/1/2#newcode435
Line 435: msg.WriteCStringToLogFile(content);
On 2009/03/31 07:23:13, Søren Gjesse wrote:
> Will this still be required if we use binary file mode?

Yes. Because the current preamble log is about 26500 bytes long, which is ten
times bigger that the current message buffer length. Either we implement
non-truncating logging or make the buffer length that big. But I think that
writing directly from the preamble log is simplier.

http://codereview.chromium.org/56064/diff/1/6
File src/platform.h (right):

http://codereview.chromium.org/56064/diff/1/6#newcode157
Line 157: static void FSetMode(FILE* stream, bool text_mode);
On 2009/03/31 07:23:13, Søren Gjesse wrote:
> I think we should try to avoid having this function. How about making sure
that
> on windows "b" is added to the mode when performing fopen on Windows. That
> should avoid any translation when writing to files on Windows. Having just \n
on
> Windows is preferable to adding this function and trying to control the file
> mode while writing.

A bright idea indeed! I checked that Python on Windows works OK with files that
have UNIX-style line ends, so I can get rid of this file mode complexity.

http://codereview.chromium.org/56064/diff/1/7
File src/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/56064/diff/1/7#newcode828
Line 828: LOG(RegExpCodeCreateEvent("RegExp", *code, *source));
On 2009/03/31 07:23:13, Søren Gjesse wrote:
> Is the first parameter required for this RegExp specific function?

Removed, thanks.

[Søren Thygesen Gjesse, 2009-03-31 09:05:07]

LGTM