595538 - r-tree benchmark gives (void 0) is undefined

Status: RESOLVED DUPLICATE of bug 541191

(Core :: JavaScript Engine, defect)

Description

[Renaud Durlin]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0b6pre) Gecko/20100911 Firefox/4.0b6pre
Build Identifier: Mozilla/5.0 (Windows NT 5.1; rv:2.0b6pre) Gecko/20100911 Firefox/4.0b6pre

Error when running a javascript benchmark.


Reproducible: Always

Steps to Reproduce:
1. go to http://stackulator.com/rtree/numbers.html

Actual Results:  
Error: (void 0) is undefined
Source File: http://stackulator.com/rtree/rtree.js
Line: 665

Expected Results:  
No error!

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Confirmed.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Seems similar to bug 539553.  Can we just add this entire benchmark to our correctness tests?

Comment 3

[David Mandelin [:dmandelin]]

Looks like this one already shows up in 4.0b5, so it is probably not JM. I think that also means it is not scope removal. Regression window on TM nightlies would be helpful.

Comment 4

[Matt]

This error, "b is undefined", started before may 2010.  Do you need where this error started?

Regression window:

Works - Error: b is undefined:
Mozilla/5.0 (Windows NT 6.1; rv:2.0b4pre) Gecko/20100810 Minefield/4.0b4pre
http://hg.mozilla.org/tracemonkey/rev/442559dabdf5

Broken - Error: (void 0) is undefined:
Mozilla/5.0 (Windows NT 6.1; rv:2.0b4pre) Gecko/20100811 Minefield/4.0b4pre
http://hg.mozilla.org/tracemonkey/rev/f84b470314a8

Pushlog:
http://hg.mozilla.org/tracemonkey/pushloghtml?fromchange=442559dabdf5&tochange=f84b470314a8

Comment 5

[David Mandelin [:dmandelin]]

(In reply to comment #4)
> This error, "b is undefined", started before may 2010.  Do you need where this
> error started?
> 
> Regression window:
> 
> Works - Error: b is undefined:
> Mozilla/5.0 (Windows NT 6.1; rv:2.0b4pre) Gecko/20100810 Minefield/4.0b4pre
> http://hg.mozilla.org/tracemonkey/rev/442559dabdf5
> 
> Broken - Error: (void 0) is undefined:
> Mozilla/5.0 (Windows NT 6.1; rv:2.0b4pre) Gecko/20100811 Minefield/4.0b4pre
> http://hg.mozilla.org/tracemonkey/rev/f84b470314a8
> 
> Pushlog:
> http://hg.mozilla.org/tracemonkey/pushloghtml?fromchange=442559dabdf5&tochange=f84b470314a8

Are you saying that "b is undefined" is correct behavior, or just using "Works" as a tag for the first rev of the regression window?

Assuming the latter, I think they are pretty much the same bug, just with a different error report. So it would be great if you could track down when the original bug started.

Comment 6

[Matt]

Yes, "Works" just designates the build found before it broke since the broken build should match the reporters description.  The errors are just listed to show how they differ.

Comment 7

[Boris Zbarsky [:bzbarsky]]

"b is undefined" is bug 541191, no?  Which I thought we fixed in bug 539553 on trunk, but apparently not.

Comment 8

[Matt]

It looks like there is a fix in works in bug 541191 so I will wait and see if a regression window is still needed.

Comment 9

[Renaud Durlin]

(In reply to comment #7)
> "b is undefined" is bug 541191, no?  Which I thought we fixed in bug 539553 on
> trunk, but apparently not.

This benchmark is broken in to different ways.

"b i undefined" appears in 3.6 and was fixed between 3.6 and 3.6.9 (I don't know the exact release) but it seems that the patch didn't land on trunk (which is strange because I thought that patch first land on trunk then go to release branch).

Then something changed the error to "(void 0) is undefined". I'm not sure if this is the same problem, this is why i filled another bug. Maybe the patch in bug 541191 also fix this bug as I don't see any error with Firefox 3.6.9.

Comment 10

[David Anderson [:dvander] - inactive, e-mail if emergency]

The "b is undefined" bug 541191 was never really fixed. It might be that the test case changed, or we're no longer tracing through the path that exposes the bug.