When looking at reftests which you're unfamiliar with, it's occasionally unclear which image is the reference and which is the failing case - sometimes they both look reasonable. It'd be great if the images had more meaningful labels.
TBPL uses (in a rather dirty way) the reftest analyser from mozilla-central: https://hg.mozilla.org/mozilla-central/file/tip/layout/tools/reftest/reftest-analyzer.xhtml (Loaded from https://hg.mozilla.org/webtools/tbpl/file/01c330f101de/js/UserInterface.js#l1510) Agree it would be good to rename them :-)
Component: Tinderboxpushlog → Reftest
Product: Webtools → Testing
Well, the use of "test" and "reference" is really a convention rather than the only way to use reftest, so I'm a little hesitant to label the images as other than "image 1" and "image 2". In a test with: == file1 file2 "image 1" is the image from file1, and "image 2" is the image from file2; it's just the order of arguments to the == operation. By "other ways to use reftest", I mean things like this: (from layout/reftests/table-width/reftest.list) == shrink-wrap-pref.html shrink-wrap-pref-ref.html == shrink-wrap-avail.html shrink-wrap-avail-ref.html == shrink-wrap-min.html shrink-wrap-min-ref.html != shrink-wrap-min-ref.html shrink-wrap-pref-ref.html != shrink-wrap-pref-ref.html shrink-wrap-avail-ref.html != shrink-wrap-avail-ref.html shrink-wrap-min-ref.html (from layout/reftests/bugs/reftest.list... though perhaps this one is bad style) == 404030-1.html 404030-1-ref.html != 404030-1-notref.html 404030-1.html != 404030-1-notref2.html 404030-1.html
That said, I guess reftest does bias towards the first file by using the first of the two arguments to identify the test.
Perhaps we could use "Test (file.html)" and "Reference (file-ref.html)". In the common case the Test/Reference labelling would be correct, but in cases where it isn't, the filenames should clear things up.
Status: NEW → RESOLVED
Last Resolved: 4 years ago
Resolution: --- → DUPLICATE
Duplicate of bug: 1016260
You need to log in before you can comment on or make changes to this bug.