Bug 16459 – REGRESSION: assertion failure with regexp with \B in a case-ignoring character range

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Darin Adler]

We get an assertion compiling this regular expression:

    /|[x\B-\u00b5]/i;

<rdar://problem/5646361>

[Darin Adler]

The issue is specific to the use of the "\B" escape sequence inside a character class. It's supposed to be an error to do that.

[Darin Adler]

Created attachment 17924 [details]
patch

[Alexey Proskuryakov]

I haven't looked at the spec yet, but in Firefox, javascript:alert(/|[x\B-\u00b5]/) doesn't raise an error. Perhaps the test and/or ChangeLog should mention that this is an intentional difference in behavior.

[Darin Adler]

(In reply to comment #3)
> I haven't looked at the spec yet, but in Firefox,
> javascript:alert(/|[x\B-\u00b5]/) doesn't raise an error. Perhaps the test
> and/or ChangeLog should mention that this is an intentional difference in
> behavior.

We could also have \B mean just "B" in character classes if that matches the other browsers. That's a simple change.

[Maciej Stachowiak]

Comment on attachment 17924 [details]
patch

Marking r- per Alexey's comment - it would be better to match other browsers for this case. (If that's hard to do right away, we can file a follow-up bug.)

[Darin Adler]

Created attachment 17942 [details]
patch, this time matching behavior of Gecko rather than the ECMAScript spec

[Darin Adler]

(In reply to comment #5)
> it would be better to match other browsers
> for this case. (If that's hard to do right away, we can file a follow-up bug.)

It was easy to do. I haven't tested IE yet, though.

[Maciej Stachowiak]

Comment on attachment 17942 [details]
patch, this time matching behavior of Gecko rather than the ECMAScript spec

r=me

[Darin Adler]

Committed revision 28785.
Committed revision 28787.