StringToDouble rewritten not using String::Get and memory allocations....

StringToDouble rewritten not using String::Get and memory allocations.

It converts the number to "canonical" form removing insignificant digits,
leading zerroes and spaces what guarantees to fit a fixed size buffer and
does not changes result of strtod.

Committed: http://code.google.com/p/v8/source/detail?r=4282

[SeRya, 2010-03-24 15:19:30]

Is is http://codereview.chromium.org/1096002/show with the following
differences:

- Fixed debug build error removing obsolete ASSERT.
- Fixed Windows build adding static_cast in conversions.cc when a new char is
copied to the buffer. Also a few changes in test-conversions.cc. 
- Fixed fest failures on 'sputnik' and 'es5conform'. Numbers like '0.', '0e1'
etc was parsed as octals. Now numbers with no significant digits in the integer
parts are parsed as decimals. Added tests for this cases into the str-to-num.js.
- The max_significant_digits constant moved out of the function and explained.

[Kasper Lund, 2010-03-24 15:40:51]

Drive by comment:

http://codereview.chromium.org/1216005/diff/1/2
File test/cctest/test-conversions.cc (right):

http://codereview.chromium.org/1216005/diff/1/2#newcode163
test/cctest/test-conversions.cc:163: #ifdef _WIN32
Why not use OS::SNPrintF instead? This is hard to read.

[SeRya, 2010-03-24 15:45:44]

http://codereview.chromium.org/1216005/diff/1/2
File test/cctest/test-conversions.cc (right):

http://codereview.chromium.org/1216005/diff/1/2#newcode163
test/cctest/test-conversions.cc:163: #ifdef _WIN32
On 2010/03/24 15:40:51, Kasper Lund wrote:
> Why not use OS::SNPrintF instead? This is hard to read.

Done.

[SeRya, 2010-03-25 13:03:45]

Fixed issues from http://codereview.chromium.org/1096002/diff/3003/39003

[Erik Corry, 2010-03-25 13:43:05]

LGTM

http://codereview.chromium.org/1216005/diff/18001/19003
File src/conversions.cc (right):

http://codereview.chromium.org/1216005/diff/18001/19003#newcode166
src/conversions.cc:166: // mean of 2 adjucent doubles (that could have up to 769
digits) the result
adjucent -> adjacent

http://codereview.chromium.org/1216005/diff/18001/19003#newcode496
src/conversions.cc:496: // Otherwise there is no digits in the string.
is no -> are no

[SeRya, 2010-03-25 14:29:49]

Changed the MaximumSignificantDigits test to avoid possible platform dependency.

http://codereview.chromium.org/1216005/diff/18001/19003
File src/conversions.cc (right):

http://codereview.chromium.org/1216005/diff/18001/19003#newcode166
src/conversions.cc:166: // mean of 2 adjucent doubles (that could have up to 769
digits) the result
On 2010/03/25 13:43:05, Erik Corry wrote:
> adjucent -> adjacent

Done.

http://codereview.chromium.org/1216005/diff/18001/19003#newcode496
src/conversions.cc:496: // Otherwise there is no digits in the string.
On 2010/03/25 13:43:05, Erik Corry wrote:
> is no -> are no

Done.