Bug 17942
| Summary: | Slickspeed native tests return 0 ms | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Brandon Shea <greatcaffeine> |
| Component: | WebCore JavaScript | Assignee: | Nobody <webkit-unassigned> |
| Status: | UNCONFIRMED | ||
| Severity: | Normal | CC: | mrowe |
| Priority: | P2 | Keywords: | InRadar |
| Version: | 528+ (Nightly build) | ||
| Hardware: | PC | ||
| OS: | Windows Vista | ||
Brandon Shea
Using WebKit's Slickspeed page (http://webkit.org/perf/slickspeed/), the latest Windows builds of WebKit do not correctly handle the native tests. The vast majority of the native tests being run simply return 0 ms for the time. Only three are handled correctly:
div:nth-child(2n)
div:nth-child(2n+1)
div:nth-child(n)
All of the non-native tests run as expected and produce results which are similar to OS X.
If it matters, this machine is running Windows Vista 64 bit, and the problem was also present on Windows XP x64 Edition.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Mark Rowe (bdash)
I wonder if this is due to the time resolution that we retrieve from Windows being less than stellar?
Mark Rowe (bdash)
<rdar://problem/5807050>
Brandon Shea
Interestingly enough, I tested this out on a 32 bit version of Windows XP, and almost all of the tests work correctly in both Safari 3.1 and the latest WebKit build. That would seem to isolate the problem to 64 bit Windows, though I'm not sure how or why that has an effect.
Geoffrey Garen
I think Steve fixed this in http://trac.webkit.org/changeset/36578.