Bug 17458

Summary:

Bug in 3d-cube SunSpider test (also in Jetstream2)

Product:

WebKit

Reporter:

John Resig <jresig>

Component:

Tools / Tests

Assignee:

Maciej Stachowiak <mjs>

Status:

NEW ---

Severity:

Normal

CC:

brendan, gavin.sharp, mjs, oliver, saam, seth.gaurav

Priority:

Version:

528+ (Nightly build)

Hardware:

All

OS:

All

Attachments:

Description	Flags
3d-cube.js patch	none

Description John Resig 2008-02-20 13:42:49 PST

There's a minor bug in the test which allows variables to leak into the global scope. Please see the attached patch.

Comment 1 John Resig 2008-02-20 13:43:31 PST

Created attachment 19238 [details]
3d-cube.js patch

Comment 2 Alexey Proskuryakov 2008-02-21 22:24:12 PST

Comment on attachment 19238 [details]
3d-cube.js patch

I assume this was meant for review; marking as such.

Comment 3 Darin Adler 2008-03-02 18:43:56 PST

Comment on attachment 19238 [details]
3d-cube.js patch

This change looks right to me. But Maciej should really be the one to review it.

Comment 4 Eric Seidel (no email) 2008-12-01 12:28:15 PST

Comment on attachment 19238 [details]
3d-cube.js patch

This bug doesn't need to sit in the review queue.  Maciej knows of the issue.  Sitting in the review queue for the next 10 months isn't going to solve the issue. :)  Maciej has noted reviewing this patch is blocked on other issues with sunspider distribution.

Comment 5 Maciej Stachowiak 2010-02-28 14:24:00 PST

This bug doesn't seem to affect the correctness of the result. Saving for SunSpider 2.0.

Comment 6 Maciej Stachowiak 2020-11-25 13:44:21 PST

Saam, can you take a look at this? Worth fixing for the next JS2 update?

Comment 7 Saam Barati 2020-11-30 16:01:12 PST

(In reply to Maciej Stachowiak from comment #6)
> Saam, can you take a look at this? Worth fixing for the next JS2 update?

I've noticed this before. And it's very possible we've done things to make this faster.

It's not quite obvious to me what the purpose of that loop is in general, especially if we switch to using `new`, none of those objects will be pointed to by anything. A sufficiently smart compiler could omit those allocations.