314465 - Implement CompareUTF8toUTF16 and zero-copy atom lookup

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Boris Zbarsky [:bzbarsky]]

Spun off of bug 277479, since we have a patch to do this and it's not obvious that that bug, as filed, should be fixed.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: jst's patch from bug 277479 updated to tip

This is updated to dbaron's, bsmedberg's, and my comments with the following exceptions:

bsmedberg:

> Do we know the string lengths before we start? Would it be quicker to hash all
> strings from end to front?

This seems kinda hard with the whole high/low surrogate thing... at least at first blush.

> Why is this inline?

It's not....

me:

> This latter is the common case, I think, and in that case we're doing ...

I realized that in the hash miss case we need to convert to UTF8 here anyway, so jst's code is all good.

shaver, this has reviews of all sorts already; we just need an xpcom moa from you here... That said, if you want to check over the changes I've made since jst's patch I'm not going to object.

Comment 2

[Mike Shaver (:shaver emeritus)]

Comment on attachment 201391 [details] [diff] [review]
jst's patch from bug 277479 updated to tip

I'll put this in my list, try to get to it on the plane tomorrow.  If not, it'll be Tuesday evening.

Comment 3

[Mike Shaver (:shaver emeritus)]

Comment on attachment 201391 [details] [diff] [review]
jst's patch from bug 277479 updated to tip

moar=shaver (sorry, I could have _sworn_ I did this on Wednesday).

Comment 4

[Boris Zbarsky [:bzbarsky]]

Fixed on trunk.  No obvious impact on our tinderbox tests, though this was helping a good bit in some other (slightly more involved) DHTML testcases I was testing on.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Note: this had bug 316573 (which has been in our codebase for ages, basically...)