[webkit-reviews] review denied: [Bug 111914] Improve string equal() speed on x86 and x86_64 : [Attachment 192326] Patch

[bugzilla-daemon at webkit.org]

Gavin Barraclough <barraclough at apple.com> has denied Benjamin Poulain
<benjamin at webkit.org>'s request for review:
Bug 111914: Improve string equal() speed on x86 and x86_64
https://bugs.webkit.org/show_bug.cgi?id=111914

Attachment 192326: Patch
https://bugs.webkit.org/attachment.cgi?id=192326&action=review

------- Additional Comments from Gavin Barraclough <barraclough at apple.com>
View in context: https://bugs.webkit.org/attachment.cgi?id=192326&action=review


Not sure I'm convinced on the safety of the byte compare – r- for now, 'til you
persuade me. :o)

> Source/WTF/wtf/text/StringImpl.h:826
> +    // at least one character in the word, so it is safe to load the full
word.

This is true, provided that the pointers 'a' & 'b' are word aligned – but it's
explicitly the case that this is not required by this method!
If the pointer is unaligned upon entry then it will still be unaligned now.  If
this function is called on an unaligned string that ends within the last word
of a page we'll read beyond page bounds and potentially fault.