REGRESSION (r66156): Sites using AppleConnect for authentication fail to log in.
Created attachment 65904 [details] Patch
<rdar://problem/8366089>
Comment on attachment 65904 [details] Patch > + bool forceFallback = frame->settings()->needsSiteSpecificQuirks() && document()->url().host() == "wdg2.apple.com"; If you put this into an inline function rather than a boolean local variable then you would not be evaluating it if beforeLoadAllowedLoad is true. In the current code you always evaluate this, even if you don't use its result. I think this needs a "why" comment in the code. It can be a brief one. Is there a guarantee that frame->settings() can't be 0? What if we are in a frame after it is no longer connected to a page? Then frame->settings() will be 0. Are we guaranteed safe from that? Seems OK, r=me
Created attachment 65980 [details] Patch
Darin, based on our conversation this morning, here is an updated patch that keys off the plug-in MIME type rather than the URL and works by lower-casing the plug-in parameters rather than falling back to the embed.
Comment on attachment 65980 [details] Patch > + if (frame && frame->settings()->needsSiteSpecificQuirks() && mimeType == "application/x-snkp") { Since MIME types are not case sensitive it would be better to use equalIgnoringCase. You’ll also need to update StringsNotToBeLocalized.txt to add this new non-localized string.
Committed http://trac.webkit.org/changeset/66437.