481738 - Consider disabling "g" 3rd argument for String.replace

Status: RESOLVED DUPLICATE of bug 1108382

(Core :: JavaScript Engine, defect)

Description

[David Mandelin [:dmandelin]]

This non-ECMA option ends up costing us about 30ms on SunSpider, because the other browsers do nothing with it, but it makes us do a global replace. 

The SunSpider benchmark has been fixed (by removing the "g") in WebKit svn, but the web version of SunSpider still has the "g", and as far as we know, it always will.

Comment 1

[Mike Shaver (:shaver emeritus)]

Are they using the known-bogus version with 'g' in their marketing materials, I wonder?

I strongly suspect that we use 'g' in our application code, so removing it isn't a great option.

Comment 2

[David Mandelin [:dmandelin]]

(In reply to comment #1)
> Are they using the known-bogus version with 'g' in their marketing materials, I
> wonder?

I don't know. I don't suppose they show detailed FF results that would enable us to check that.

Is there any way we can persuade them to remove it? I know they want to keep benchmarks comparable over time, but as this is a flaw in the benchmark and it will only change the score of FF, that justification seems not to apply.

Comment 3

[Tooru Fujisawa [:arai]]

it's tracked in bug 1108382, and now in-tree consumers are removed from m-c and c-c, I'm going to add warning.

Comment 4

[Tooru Fujisawa [:arai]]

now I think this is simply a dupe of bug 1108382.