Summary: | SunSpider computes incorrect two-sample t statistics | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Adam Roben (:aroben) <aroben> | ||||
Component: | New Bugs | Assignee: | Adam Roben (:aroben) <aroben> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | eric, mjs | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Other | ||||||
OS: | All | ||||||
Attachments: |
|
Description
Adam Roben (:aroben)
2010-05-07 12:49:48 PDT
Created attachment 55410 [details]
Patch
Attachment 55410 [details] was posted by a committer and has review+, assigning to Adam Roben for commit.
I wonder if we need to somehow communicate that SunSpider was buggy in this way? I guess not, since the t statistics (or any values derived from them) don't get persisted in the run results that people might save for the future; they're only used when actually comparing two runs. (In reply to comment #3) > I wonder if we need to somehow communicate that SunSpider was buggy in this way? I guess not, since the t statistics (or any values derived from them) don't get persisted in the run results that people might save for the future; they're only used when actually comparing two runs. In fact, my patch doesn't change the versions of SunSpider hosted on webkit.org, so this is definitely not necessary at the moment. I wonder if we should change those versions, though? Committed r59370: <http://trac.webkit.org/changeset/59370> Maciej, do you think we should apply this change to the public versions of SunSpider? |