246911 - misbehave of switch() statement with variable as label

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Roman Ivanov]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.6) Gecko/20040206 Firefox/0.8
Build Identifier: Firefox/0.9

This code:
var a = 10;
a = 9;
var b = 10;
switch(b)
{
 case a:
  alert('got it');
  break;
 default:
  alert('passed');
}
give 'passed' in all browsers including MSIE & FireFox 0.8, but gives 'got it'
under FireFox 0.9.

As it seems to me, it's a bug.

Reproducible: Always
Steps to Reproduce:
1. Write your code with switch() statement using variables as labels.

Actual Results:  
Variable are always constant in switch() statement.

Expected Results:  
Variable must change its value when I change it.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: reporter's testcase

Comment 2

[Brendan Eich [:brendan]]

Something bad in the browser embeddings.  We can't reproduce this in the shells,
AFAIK.

/be

Comment 3

[Johnny Stenback (:jst)]

Hmm, trunk builds pass too, is this aviary branch only?

Comment 4

[Johnny Stenback (:jst)]

Hmm, if I copy the JS engine source into my aviary branch tree and rebuild, the
bug goes away. Did something get fixed on the trunk and not on the aviary
branch, or was something incorrectly merged?

Comment 5

[Scott MacGregor]

Here are the 1.7 bugs that were merged into the aviary branch:

http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=MozillaTinderboxAll&branch=AVIARY_1_0_20040515_BRANCH&branchtype=match&dir=mozilla%2Fjs&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=month&mindate=&maxdate=&cvsroot=%2Fcvsroot

I'd look here first as this is one patch I had to apply by hand:
 
Bug #245795

Comment 6

[Johnny Stenback (:jst)]

This happens on the 1.7 branch as well, so unlikely a aviary branch merge problem.

Comment 7

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Found it.  Patch coming up now.

Comment 8

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Attachment: only define compile-time constants for JSOP_DEFCONST

Looking for quick r/sr for "obvious" fix.  Risk very limited: might cause us to
miss a chance to optimize a constant, at absolute worst.

Comment 9

[Brendan Eich [:brendan]]

Comment on attachment 150869 [details] [diff] [review]
only define compile-time constants for JSOP_DEFCONST

Curse me for sneaking a fix into a big landing.  I had completely forgotten
about this!

/be

Comment 10

[Brendan Eich [:brendan]]

In case it's unclear, this fix has been on the trunk a while.  It's good for the
aviary branch too; I think Ben is syncing up.  Someone slap the right keywords
or status whiteboard on this bug and resolve it.

/be

Comment 11

[Michael Lefevre]

looks like this has been checked in to 1.7 and aviary now, so slapping as requested.

Comment 12

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

We need a test case for this in the suite, I beseech thee!

Comment 13

[Tracy Walker [:tracy]]

using the reporters test case with 1.7, this is verified.

Comment 14

[Bob Clary [:bc] (inactive)]

Attachment: js1_5/Regress/regress-246911.js

Roman, with your permission this will be included in the javascript test
library.

Comment 15

[Roman Ivanov]

> Roman, with your permission this will be included in the javascript test
> library.

Yes, of course.

Thank you very much.

Comment 16

[Bob Clary [:bc] (inactive)]

js1_5/Regress/regress-246911.js checked in.