InternalSettingsGenerated::resetToConsistentState() may override TestOptions such as enableBackForwardCache. This is causing issues such as the back/forward cache sometimes getting enabled on tests where TestOptions::enableBackForwardCache() is false, because a previous test has enabled the back/forward cache.
Created attachment 390265 [details] Patch
Comment on attachment 390265 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=390265&action=review > Tools/DumpRenderTree/mac/DumpRenderTree.mm:1927 > + resetWebPreferencesToConsistentValues(); > + setWebPreferencesForTestOptions(options); I recall that we had bugs fixed by moving various "reset preferences" calls around in the past too. Probably not these exact ones. Can you add a comment explaining the invariant that need to be maintained ("do this after ... because ...")? That will help keep this fix preserved for longer. And maybe even a pair of regression tests that reproduces this directly and reliably?
Created attachment 390307 [details] Patch
Comment on attachment 390307 [details] Patch Clearing flags on attachment: 390307 Committed r256232: <https://trac.webkit.org/changeset/256232>
All reviewed patches have been landed. Closing bug.
<rdar://problem/59331661>
Reverted r256232 for reason: Caused ResizeObserver tests to become flaky Committed r258438: <https://trac.webkit.org/changeset/258438>
Created attachment 393544 [details] Patch
Created attachment 393550 [details] Patch
Comment on attachment 393550 [details] Patch Clearing flags on attachment: 393550 Committed r258482: <https://trac.webkit.org/changeset/258482>