Bug 21414 – REGRESSION: Regular Expressions and character classes, shorthands and ranges

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

[Bernd Lutz]

JavaScript has some difficulties with character classes and shorthands like "\w" while using a hyphen as hyphen in a range.

JS example: 'hello_there'.replace(/[\w\-]+/, 'xxx')
This should return the string "xxx" like in Safari 3.1.2 or Firefox 3.x. The current WebKit Build returns "xxx_there" because the character class shorthand \w seems not to include the underscore anymore. Removing the "\-" of the range fixes the bug.

For some reasons you cannot simply place a underscore within a character class to fix the erratic behaviour related to the \w shorthand. The underscore must be escaped (this is normally not necessary).

[Alexey Proskuryakov]

Giving this full regression treatment. On IRC, it was mentioned that the issue is still present in current nightlies (I haven't tried to reproduce it).

[Gianni Chiappetta]

Created attachment 27183 [details]
This is a test case for the bug which provides an example.

Provides a quick example of the bug with actual results and expected results.

[Gianni Chiappetta]

I've come across this bug while testing scripts in the latest nightly. I've added a test case to illustrate the bug.

[Alexey Proskuryakov]

Confirmed with r40407.

[Alexey Proskuryakov]

<rdar://problem/6543487>

[Oliver Hunt]

Committing to http://svn.webkit.org/repository/webkit/trunk ...
	M	JavaScriptCore/ChangeLog
	M	JavaScriptCore/wrec/WRECGenerator.cpp
	M	LayoutTests/ChangeLog
	A	LayoutTests/fast/js/regexp-character-match-out-of-order-expected.txt
	A	LayoutTests/fast/js/regexp-character-match-out-of-order.html
	A	LayoutTests/fast/js/resources/regexp-character-match-out-of-order.js
Committed r40522