is a constant text failure on iOS 14.
-PASS: Correct Selection
-PASS: Correct Selection
+FAIL: failed to select a word as a result of a long press. Incorrect Selection:
+FAIL: failed to reduce selection to a single character by dragging down. Incorrect Selection:
This test started to constantly fail at r281054: https://trac.webkit.org/changeset/281054/webkit
It looks like changes were meant for iOS15, but have also had an impact on iOS14 testing.
It's surprising to me that we attempt to run this test in OpenSource bots, considering that it's under fast/events/touch/ios. That said, this test doesn't rely on touch events, that's not necessarily a bad thing.
> It looks like changes were meant for iOS15, but have also had an impact on
> iOS14 testing.
The changes were not specifically targeted for iOS 14 or 15.
(In reply to Wenson Hsieh from comment #3)
> > It looks like changes were meant for iOS15, but have also had an impact on
> > iOS14 testing.
> The changes were not specifically targeted for iOS 14 or 15.
That's good to know. I haven't dug too far into the details for this yet. My assumption for the targeting was because the bug that introduced this failure was titled with [ iOS 15 ]
Oh, I think I know what's going on.
After r281054, `fast/events/touch/ios/long-press-then-drag-to-select-text.html` is now the first test that's run in OpenSource inside the `fast/events/touch/ios` directory. There's probably a latent bug that causes `long-press-then-drag-to-select-text.html` to fail if it's run using a newly-created WKWebView.
My patch made us stop running `fast/events/touch/ios/long-press-on-link.html` in OpenSource, since that test was marked as PASS for iOS 14 and I removed that PASS expectation (along with the more broad CRASH expectation for all iOS).
This means we fall back to the directory-level expectations in `fast/events/touch/ios`, which is skipped in OpenSource. Prior to r281054, this test was the first in its directory; now, long-press-then-drag-to-select-text.html is the first test.
I suspect that *restoring* the PASS expectation for `long-press-on-link.html` will "fix" this test failure by making long-press-on-link.html the first test in its directory again. It seems okay to do this too, because `long-press-on-link.html` is a test that we eventually do want to run in open source bots, because it doesn't depend on touch events.
Committed r281103 (240562@main): <https://commits.webkit.org/240562@main>
Let's see if that "fixes" it.
(In reply to Wenson Hsieh from comment #6)
> Committed r281103 (240562@main): <https://commits.webkit.org/240562@main>
> Let's see if that "fixes" it.
https://results.webkit.org/?suite=layout-tests&test=fast%2Fevents%2Ftouch%2Fios%2Flong-press-then-drag-to-select-text.html is starting to show successful runs.