Migrate some Apple internal Mac expectations here into WebKit sources
Created attachment 406531 [details] Patch
<rdar://problem/66161781>
Created attachment 406532 [details] Patch
Comment on attachment 406532 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=406532&action=review > LayoutTests/platform/mac-wk2/TestExpectations:1211 > +[ arm64 ] http/tests/security/storage-blocking-loosened-plugin.html [ Skip ] As suggested by Alexey, can and will break this into two lines to get rid of the run-webkit-tests lint warning: [ arm64 Release ] http/tests/security/storage-blocking-loosened-plugin.html [ Skip ] [ arm64 Debug ] http/tests/security/storage-blocking-loosened-plugin.html [ Skip ]
This is associated with the wrong bug in Radar. How do I fix that?
(In reply to Darin Adler from comment #5) > This is associated with the wrong bug in Radar. How do I fix that? Paste the rdar URL in a new comment and I believe the importer should take care of it.
rdar://67026638
Looking for review on this; it will be easier to merge if itβs done before people make a lot of other changes to test expectations.
Comment on attachment 406532 [details] Patch Looks like our flakey test is not one we had an expectation change for, so let's land!
Committed r265639: <https://trac.webkit.org/changeset/265639>
Comment on attachment 406532 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=406532&action=review This is a very large patch that mixes different kinds of changes. I think that some of these are desirable, but not all. Only did a spot check, as this is so large, and already landed. > LayoutTests/platform/mac-wk1/TestExpectations:1087 > +# <rdar://problem/61461935> [ macOS wk2 ] fast/scrolling/rtl-point-in-iframe.html is flaky failing > +[ Mojave ] fast/scrolling/rtl-point-in-iframe.html [ Pass Failure ] This isn't failing anywhere, open source or internal. > LayoutTests/platform/mac/TestExpectations:1944 > +# <rdar://problem/31017631> Stress GC bot failing LayoutTest step with timeout while running WebCryptoAPI/derive_bits_keys tests > +[ Debug ] imported/w3c/web-platform-tests/WebCryptoAPI/derive_bits_keys [ Skip ] How is this a reason to skip the test on all Mac bots? > LayoutTests/platform/mac/TestExpectations:1948 > +# The navigator.language tests rely on functionality only available in recent releases of macOS Catalina and onward. > +# <rdar://problem/59461792> REGRESSION: [ Mac iOS ] ( r254389 - r254411 ) fast/text/international/system-language/declarative-language.html is failing internally > +fast/text/international/system-language/declarative-language.html [ Pass Failure ] The expectation doesn't match the comment, and neither matches the actual behavior. > LayoutTests/platform/mac/TestExpectations:1955 > +# <rdar://problem/59636115> REGRESSION: [ iOS & macOS ] two imported/w3c/web-platform-tests/WebCryptoAPI are failing > +imported/w3c/web-platform-tests/WebCryptoAPI/import_export/rsa_importKey.https.worker.html [ Pass Failure ] > +imported/w3c/web-platform-tests/WebCryptoAPI/import_export/test_rsa_importKey.https.html [ Pass Failure ] This is an OS specific regression, so marking the tests flaky unconditionally regresses test coverage. > LayoutTests/platform/mac/TestExpectations:2041 > +# <rdar://problem/61114827> [ Stress GC ] http/tests/navigation/page-cache-fontfaceset.html is flaky crashing. > +[ Mojave ] http/tests/navigation/page-cache-fontfaceset.html [ Pass Crash ] All of these are a very internal hack, enabled by running Stress GC on Mojave only. This is seems confusing and misplaced in open source expectations. > LayoutTests/platform/mac/TestExpectations:2203 > +# rdar://66802945 (REGRESSION (20A2323e-20A234b): [ Layout Test] 12 canvas layout tests are a constant failure) > +[ arm64 ] canvas/philip/tests/2d.line.miter.obtuse.html [ Failure ] > +[ arm64 ] canvas/philip/tests/2d.path.arc.selfintersect.1.html [ Failure ] > +[ arm64 ] canvas/philip/tests/2d.path.arc.shape.3.html [ Failure ] > +[ arm64 ] fast/canvas/canvas-ellipse-zero-lineto.html [ Failure ] > +[ arm64 ] imported/w3c/canvas/2d.line.miter.obtuse.html [ Failure ] > +[ arm64 ] imported/w3c/canvas/2d.path.arc.selfintersect.1.html [ Failure ] > +[ arm64 ] imported/w3c/canvas/2d.path.arc.shape.3.html [ Failure ] > +[ arm64 ] imported/w3c/web-platform-tests/html/canvas/element/line-styles/2d.line.miter.obtuse.html [ Failure ] > +[ arm64 ] imported/w3c/web-platform-tests/html/canvas/element/path-objects/2d.path.arc.selfintersect.1.html [ Failure ] > +[ arm64 ] imported/w3c/web-platform-tests/html/canvas/element/path-objects/2d.path.arc.shape.3.html [ Failure ] As these are due to an issue below WebKit, moving the expectations to open source makes them harder to keep track of, I think.
I can see that you and I have very different ideas about open source vs. internal. As far as I am concerned, only secret changes should be internal.
I agree with this as a basic philosophy! What I don't believe in is that mechanically moving everything into open source is the right first step. This is an essentially non-reversible move that loses information. Tracking issues that are below WebKit and are only affecting future OSes, possibly for a short period of time that isn't aligned with public beta schedule, is somewhat of a gray area for me.
Comment on attachment 406532 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=406532&action=review > LayoutTests/ChangeLog:23 > + * platform/mac-bigsur-wk1/editing/pasteboard/drop-text-without-selection-expected.txt: Added. We all missed the big issue. Expectations for latest macOS live in platform/mac (because that's the present and also the likely future), and only older OSes get named directories. Will get these updated.