There is a number of orphaned entries in TestExpectations after the recent re-sync in r217225. The import process should make sure that expectations don't get broken. --lint-test-files warnings: LayoutTests/TestExpectations:752 Path does not exist. imported/w3c/web-platform-tests/html/webappapis/system-state-and-capabilities/the-navigator-object/NavigatorID.html LayoutTests/TestExpectations:801 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/sign_verify/test_ecdsa.html LayoutTests/TestExpectations:1021 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/derive_bits_keys/test_pbkdf2.html LayoutTests/TestExpectations:1024 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/generateKey/test_successes_RSA-OAEP.html LayoutTests/TestExpectations:1026 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/import_export/test_rsa_importKey.html LayoutTests/TestExpectations:1032 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/generateKey/test_failures.html LayoutTests/TestExpectations:1033 Path does not exist. imported/w3c/web-platform-tests/WebCryptoAPI/generateKey/test_successes.html
Thanks for filing this issue, I'll check that for next resync manually. Maybe the script could help although the current TestExpectations format is not very adequate for scripts.
The script could prompt to fix these manually if the lint output changes.
<rdar://problem/113394741>
(In reply to Alexey Proskuryakov from comment #2) > The script could prompt to fix these manually if the lint output changes. check-webkit-style usually prompts me when I leave orphaned paths, but maybe it ignores the files that haven't been edited.
Renames are a challenge, because it's rather hard to determine what is _actually_ a rename. For example we have nearly 4000 files whose content is: <!-- This file is required for WebKit test infrastructure to run the templated test --> If we end up deleting one such file and adding a new one, that almost certainly isn't actually a rename. We can certainly do better than "nothing", but we will likely have to live with false-positives (because two tests can differ in terms of what support files get loaded). What we can definitely do correctly is remove lines where the test is deleted.