202564 - JavaScript regexp matches same text in two places

Status: VERIFIED DUPLICATE of bug 85721

(Core :: JavaScript Engine, defect)

Description

[Eric Hodel]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4b) Gecko/20030411
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4b) Gecko/20030411

The Regex in the URL has 'Buckley' in the 4th group, when it should be null. 
Both IE (same code) and Ruby get this right (slightly modified code).

Reproducible: Always

Steps to Reproduce:
1. Click URL


Actual Results:  
'Buckley' is displayed in an alert box

Expected Results:  
nothing should be displayed in the alert box (no text)

Comment 1

[Phil Schwartau]

var re  = /(?:(.+), )?(.+), (..) to (?:(.+), )?(.+), (..)/
var str = "Seattle, WA to Buckley, WA"
var arr = re.exec(str);

According to Perl, the correct match array should be:

arr[0] = "Seattle, WA to Buckley, WA" 
arr[1] = ""
arr[2] = "Seattle"
arr[3] = "WA"
arr[4] = ""
arr[5] = "Buckley"
arr[6] = "WA"


SUMMARY using the shorthand |str| for the 0th match elt

Perl      [str, "", "Seattle", "WA", "", "Buckley", "WA"]
IE6       [str, "", "Seattle", "WA", "", "Buckley", "WA"]
JS 85721  [str, , "Seattle", "WA", , "Buckley", "WA"]
JS tip    [str, "", "Seattle", "WA", "Buckley", "Buckley", "WA"]


"JS 85721" refers to the JS tip with the patch for bug 85721 applied.
Note that the empty-string matches one gets in Perl and IE6 match as
|undefined| here instead of the empty string. This is because in ECMA,
regexp backreferences must hold |undefined| if not used. See bug 123437.

So modulo this difference in convention between Perl/IE and ECMA/JS,
this is one more bug that will be fixed in the big RegExp rewrite
underway in bug 85721. 

I will create a testcase from this example, add it to the testsuite,
and dupe this bug against bug 85721.

Comment 2

[Phil Schwartau]

Testcase added to JS testsuite:

      mozilla/js/tests/ecma_3/RegExp/regress-202564.js

Currently failing in the JS tip, but passing in the JS 85721 shell
and in Rhino, where the patch for bug 85721 has already gone in.

*** This bug has been marked as a duplicate of 85721 ***

Comment 3

[Phil Schwartau]

Marking Verified. 

Eric, thank you for this report. You have been cc'ed on the other
bug so you can follow progress on this issue -