First step in refactoring the logging system to allow the logging messages to...

First step in refactoring the logging system to allow the logging messages to have other destinations than an open file.

This change introduces a log message builder which builds the log message in a static string buffer instead of formatting the log message through fprintf. Currently this message can only be written to a file afterwards.

Changed the code related events to use the log message builder.

Committed: http://code.google.com/p/v8/source/detail?r=1284

[Søren Thygesen Gjesse, 2009-02-16 15:38:49]

I did not change all the logging methods to use this new abstraction as I would
like some comments first.

[Mads Ager (chromium), 2009-02-17 07:48:08]

LGTM

Would it make sense to add assertions to verify in debug mode that we never
generate log messages larger than kMessageBufferSize?

http://codereview.chromium.org/20406/diff/1001/8
File src/log.cc (right):

http://codereview.chromium.org/20406/diff/1001/8#newcode268
Line 268: virtual ~LogMessageBuilder() { }
Are you planning to use this as a base class for different builders?  If not,
there is no need for a virtual destructor.

http://codereview.chromium.org/20406/diff/1001/8#newcode296
Line 296: pos_ += result;
You need to check for a -1 result here.  VSNPrintF guards against buffer
overflow and returns -1 if the message was truncated.  In that case pos_ should
not be decremented by one (since that could lead to a negative pos_ value if
pos_ is zero).

http://codereview.chromium.org/20406/diff/1001/9
File src/log.h (right):

http://codereview.chromium.org/20406/diff/1001/9#newcode239
Line 239: // mutex_ should be acquired Before using logfile_.
Before -> before.

[Søren Thygesen Gjesse, 2009-02-17 08:20:53]

Added assertion

  ASSERT(pos_ <= Logger::kMessageBufferSize);

to LogMessageBuilder both when building the message and when writing it the
file.

http://codereview.chromium.org/20406/diff/1001/8
File src/log.cc (right):

http://codereview.chromium.org/20406/diff/1001/8#newcode268
Line 268: virtual ~LogMessageBuilder() { }
On 2009/02/17 07:48:08, Mads Ager wrote:
> Are you planning to use this as a base class for different builders?  If not,
> there is no need for a virtual destructor.

Removed virtual. I put it there to follow the new style guide where this would
fail lint.

http://codereview.chromium.org/20406/diff/1001/8#newcode296
Line 296: pos_ += result;
On 2009/02/17 07:48:08, Mads Ager wrote:
> You need to check for a -1 result here.  VSNPrintF guards against buffer
> overflow and returns -1 if the message was truncated.  In that case pos_
should
> not be decremented by one (since that could lead to a negative pos_ value if
> pos_ is zero).

pos_ is now set to kMessageBufferSize if VSNPrintF truncates output and returns
-1.

http://codereview.chromium.org/20406/diff/1001/9
File src/log.h (right):

http://codereview.chromium.org/20406/diff/1001/9#newcode239
Line 239: // mutex_ should be acquired Before using logfile_.
On 2009/02/17 07:48:08, Mads Ager wrote:
> Before -> before.

Done.