Bug 56077 – ES5 conformance issues with RegExp.prototype

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

[Gavin Barraclough]

There are three issues causing test failures in sputnik.

(1) lastIndex should be converted at the point it is used, not the point it is set (this is visible if valueOf is overridden).
(2) The 'length' property of the test/exec functions should be 1.
(3) If no input is specified, the input to test()/exec() is "undefined" (i.e. ToString(undefined)) - not RegExp.input.

[Gavin Barraclough]

Created attachment 85297 [details]
The patch (no perf impact)

[WebKit Review Bot]

Attachment 85297 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'LayoutTests/ChangeLog', u'LayoutTests/fast..." exit_code: 1

Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:55:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:56:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:60:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:61:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:70:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:71:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:75:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:76:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:80:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:81:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:85:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input_as_array.js:86:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:55:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:56:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:60:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:61:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:70:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:71:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:75:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:76:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:80:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:81:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:85:  Line contains tab character.  [whitespace/tab] [5]
Source/JavaScriptCore/tests/mozilla/js1_2/regexp/RegExp_input.js:86:  Line contains tab character.  [whitespace/tab] [5]
Total errors found: 24 in 18 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Gavin Barraclough]

Fixed in r80743

[Sergio Villar Senin]

After this patch landed we're getting some JSC crashes in GTK+ 32bit Release bot. In particular these two are failing in bots

- fast/profiler/dead-time.html
- fast/profiler/stop-profiling-after-setTimeout.html

[Gavin Barraclough]

(In reply to comment #4)
> After this patch landed we're getting some JSC crashes in GTK+ 32bit Release bot. In particular these two are failing in bots
> 
> - fast/profiler/dead-time.html
> - fast/profiler/stop-profiling-after-setTimeout.html

I think it is unlikely those crashes are related to this patch.  the GTK bots have been particularly prone to some sporadic bugs caused by GC changes lately, which hopefully olliej has all fixed now.  If you are continuing to see problems, please file a new bug.

thanks!, G.

[Sergio Villar Senin]

(In reply to comment #5)
> (In reply to comment #4)
> > After this patch landed we're getting some JSC crashes in GTK+ 32bit Release bot. In particular these two are failing in bots
> > 
> > - fast/profiler/dead-time.html
> > - fast/profiler/stop-profiling-after-setTimeout.html
> 
> I think it is unlikely those crashes are related to this patch.  the GTK bots have been particularly prone to some sporadic bugs caused by GC changes lately, which hopefully olliej has all fixed now.  If you are continuing to see problems, please file a new bug.
> 
> thanks!, G.

Well maybe that patch is not the problem and it discovers some other bug, but the thing is that if we revert r80743 locally then the crashes are gone. With it we can reproduce them always.

[Gavin Barraclough]

(In reply to comment #6)
> Well maybe that patch is not the problem and it discovers some other bug, but the thing is that if we revert r80743 locally then the crashes are gone. With it we can reproduce them always.

GTK 32-bit release doesn't seem to be building today.  Do have any stack traces of the crashes you could attach?

cheers,
G.

[Martin Robinson]

Gustavo, is it possible for you to run these tests on your bot and get a backtrace for the crash?