fast/forms/legend-access-key.html is failing since it uses "altKey" as the modifier argument. Non-array modifier arguments should be handled in EventSendingController::keyDown to solve the issue.
Created attachment 164104 [details] Patch
Comment on attachment 164104 [details] Patch Clearing flags on attachment: 164104 Committed r128602: <http://trac.webkit.org/changeset/128602>
All reviewed patches have been landed. Closing bug.
> Non-array modifier arguments should be handled in EventSendingController::keyDown to solve the issue. Are you saying that DumpRenderTree supports this, so WebKitTestRunner should, too? I don't see that in DumpRenderTree implementation. I'd like to roll out this patch unless there is a convincing explanation.
(In reply to comment #4) > > Non-array modifier arguments should be handled in EventSendingController::keyDown to solve the issue. > > Are you saying that DumpRenderTree supports this, so WebKitTestRunner should, too? I don't see that in DumpRenderTree implementation. This support was added by different ports for DumpRenderTree in EventSender, so I thought it make sense to have it in WKTR too. http://trac.webkit.org/browser/trunk/Tools/DumpRenderTree/efl/EventSender.cpp#L256 http://trac.webkit.org/browser/trunk/Tools/DumpRenderTree/gtk/EventSender.cpp#L297 http://trac.webkit.org/browser/trunk/Tools/DumpRenderTree/chromium/TestRunner/EventSender.cpp#L192
CC'ing people who were adding this code to ports over the years. This is just wrong. This was never meant to work with string arguments, and still doesn't in Mac DumpRenderTree. Was everyone just adding this code to accommodate for a typo in a single test?! How did that pass review multiple times? Please at the very least implement this on Mac too, to regain interoperability of DumpRenderTree implementations.
Seems like a reasonable feature to me. I certainly don't mind supporting this variant.
Implemented this on Mac too, in the bug #97805