Summary: | Reftests can't have platform specific results | ||
---|---|---|---|
Product: | WebKit | Reporter: | Emil A Eklund <eae> |
Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
Status: | NEW --- | ||
Severity: | Normal | CC: | abucur, ap, dpranke, leviw, zan |
Priority: | P2 | Keywords: | NRWT |
Version: | 528+ (Nightly build) | ||
Hardware: | Unspecified | ||
OS: | Unspecified |
Description
Emil A Eklund
2012-05-22 15:40:07 PDT
Is this something that needs to be supported? This seems like the whole idea of reftests. Perhaps one case where this is needed is when a feature is not implemented or disabled on some platforms, yet we don't want to disable the test for some reason. (In reply to comment #1) > Is this something that needs to be supported? This seems like the whole idea of reftests. > > Perhaps one case where this is needed is when a feature is not implemented or disabled on some platforms, yet we don't want to disable the test for some reason. This is precisely the case. We ran into this issue when enabling sub-pixel layout on Chromium. That feature can lead to contents of a page shifting by several pixels. Since the actual layout of the reference vs the test often differ, it can be extremely difficult to tweak the expectations to be identical to the test with both the feature enabled and disabled. I believe this is now supported, for instance: http://trac.webkit.org/browser/trunk/LayoutTests/fast/media/mq-transform-02-expected.html http://trac.webkit.org/browser/trunk/LayoutTests/platform/gtk/fast/media/mq-transform-02-expected.html Works as expected, the GTK-specific baseline overriding the generic one. |