https://bugs.webkit.org/show_bug.cgi?id=217301
Reviewed by Sam Weinig.
These changes do not change what the tests are testing. They remove
dependencies in these tests on "canonicalization" of selection points.
This is often always a change to the setup for the test, not the test substance.
In other cases, tests that are not trying to test selection are leaving unimportant
remnants of the selection in their results: clear the selection for those.
Patch also includes a few other expectations fixes so all the tests in the editing
directory pass for me on my development machine.
* editing/pasteboard/copy-null-characters-expected.txt: Remove expected selection.
* editing/pasteboard/copy-null-characters.html: Clear the selection after one part
of this test so the markup dump doesn't contain selection endpoints.
* editing/pasteboard/paste-noscript-xhtml-expected.txt: Remove expected selection.
* editing/pasteboard/paste-noscript-xhtml.xhtml: Clear the selection at two points
in this test so the markup dump doesn't contain selection endpoints.
* editing/selection/extend-by-sentence-002.html: Check the selected string rather
than checking the DOM node and offset of the selection. Keeps the substance of this
test without depending on the selection object selecting differently than what we
tell it to do.
* editing/selection/extend-selection-enclosing-block-mac.html: Set the position
at the start of a text node instead of before the text node.
* editing/selection/extend-selection-enclosing-block-win.html: Ditto.
* editing/selection/home-end.html: Added a setPositionAfterLeadingWhitespace
function and use it to set the position at the point the test results are
expecting. Before this change, the code relied on the Selection API to adjust the
selection to the first non-whitespace character, even though that is both non-standard
and not what we are testing. Doing this lets us keep the expected test results the same.
* editing/selection/move-by-word-visually-multi-line-expected.txt: Updated to
expect moving across the word, even the one that was preceded by a collapsed newline.
* editing/selection/move-by-word-visually-multi-line.html: Ditto.
* editing/selection/resources/extend-selection.js:
(setPositionAfterLeadingWhitespace): Added.
(runSelectionTestsWithGranularity): Use setPositionAfterLeadingWhitespace here
for the same purpose as in home-end.html above, to preserve the existing tests
without relying on non-standard Selection API behavior.
* editing/selection/resources/move-by-word-visually.js:
(setPositionAfterCollapsedLeadingWhitespace): Added.
(moveByWordForEveryPosition): Same as above, use the function.
* editing/selection/setBaseAndExtent-revert-selection.html: Fix the output function
to properly write out the node ID when the node is a text node within the node with
the ID. Before, the test could fail if the anchor node returned the text node we
passed in, but that relies on non-standard Selection API changes to the passed-in node.
* editing/selection/toString-1.html: Write out the unexpected result when the test
fails. It's not failing, but at one point it was, and this made it easier to diagnose.
* editing/undo/undo-after-event-edited.html: Use the Delete command to do a deletion
that we later undo. The old code was using the Selection API extractContents method,
which is not an undoable editing operation.
* platform/gtk/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
* platform/ios/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
* platform/mac-bigsur-wk1/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
* platform/mac-bigsur/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
* platform/win/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
* platform/wincairo/editing/pasteboard/drop-text-without-selection-expected.txt: Removed.
These unused files were incorrectly left behind when we converted this into a reftest.
* platform/mac-wk1/TestExpectations: Changed retro-correction-spelling-markers.html
to "[ Pass Failure ]" since it's a flaky test, not a reliable failure.
* platform/mac/TestExpectations: Changed some flakiness expectations that were marked
Release-only to be in effect in Debug builds too. For autocorrection-contraction.html,
expect a timeout on BigSur+ Debug. Not sure exactly what is causing it, but it's happening
both locally for me and on the bots.
* platform/wk2/TestExpectations: Expect a timeout in Debug builds for two tests that are
timing out both for me and on the Modern WebKit Debug bots, but not Legacy WebKit and not
Release bots.
Canonical link: https://commits.webkit.org/230100@main
git-svn-id: https://svn.webkit.org/repository/webkit/trunk@267997 268f45cc-cd09-0410-ab3c-d52691b4dbfc
0c4c7fcef4