keyPath of createIndex() must not be nullable (http://dvcs.w3.org/hg/IndexedDB/raw-file/tip/Overview.html#widl-IDBObjectStore-createIndex-IDBIndex-DOMString-name-DOMString-keyPath-Object-optionalParameters). However, LayoutTests/storage/indexeddb/mozilla/create-index-null-name.html is trying to test that null is accepted as the keyPath of createIndex().
Created attachment 99663 [details] Patch
If WebKit behavior is incorrect, then it should be fixed, and regression test should be corrected to test for proper behavior. How does removing a test help?
> If WebKit behavior is incorrect, then it should be fixed, and regression test should be corrected to test for proper behavior. How does removing a test help? I am trying to fix a keypath parser in WebKit bug 62288, and then found that storage/indexeddb/mozilla/create-index-null-name.html fails with my patch. So I thought that we should first remove the test (because the test is against the spec) as another bug. If you think that I should remove the test in bug 62288, I would like to just close this bug and do it in bug 62288.
Yes, we regularly make changes to code and tests in the same patch. That minimizes confusion when looking at the changes later (for example to study them, to revert or to merge on a branch).
In addition to keeping the test changes with the code changes, can we update the test to reflect the correct behavior rather than removing it completely?
> I would like to just close this bug and do it in bug 62288. So, let's do that.
(In reply to comment #5) > In addition to keeping the test changes with the code changes, can we update the test to reflect the correct behavior rather than removing it completely? Yes, I am doing this in bug 62288. In bug 62288, I removed create-index-null-name.html and instead added the test that reflects the correct behavior to keypath-basics.html. Closing this bug.