imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders-cl.htm imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders.htm imported/w3c/web-platform-tests/XMLHttpRequest/setrequestheader-content-type.htm All of them fail because libsoup fails to load the requests because the response doesn't include the empty line after the headers. Other browsers allow that, so I've written a patch for libsoup in https://bugzilla.gnome.org/show_bug.cgi?id=780352. While the patch is reviewed and we have a new libsoup version to depend on, we can probably add the patch to our jhbuild to make sure those test pass and don't regress.
Actually, with libsoup patched imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders-cl.htm passes, imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders.htm still fails, but it no longer timeouts and resultds are pretty much the same, and imported/w3c/web-platform-tests/XMLHttpRequest/setrequestheader-content-type.htm still fails with the same errors.
imported/w3c/web-platform-tests/XMLHttpRequest/setrequestheader-content-type.htm no longer works with the current version of wptserve, so I'm going to remove our expectations, since the test is skipped globally.
Our result for imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders.htm is valid with libsoup patched, it's just different to what CF does, but still valid.
Created attachment 305002 [details] Patch
Comment on attachment 305002 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=305002&action=review > LayoutTests/platform/gtk/imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders-expected.txt:3 > +FAIL XMLHttpRequest: getAllResponseHeaders() 1 assert_equals: expected "also-here: Mr. PB\r\newok: lego\r\nfoo-test: 1, 2\r\n" but got "ALSO-here: Mr. PB\r\nfoo-test: 1, 2\r\newok: lego\r\n" IMO the test should be updated to print PASS for either result, so we don't have the text FAIL in our expected results unnecessarily.
(In reply to Michael Catanzaro from comment #5) > Comment on attachment 305002 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=305002&action=review > > > LayoutTests/platform/gtk/imported/w3c/web-platform-tests/XMLHttpRequest/getallresponseheaders-expected.txt:3 > > +FAIL XMLHttpRequest: getAllResponseHeaders() 1 assert_equals: expected "also-here: Mr. PB\r\newok: lego\r\nfoo-test: 1, 2\r\n" but got "ALSO-here: Mr. PB\r\nfoo-test: 1, 2\r\newok: lego\r\n" > > IMO the test should be updated to print PASS for either result, so we don't > have the text FAIL in our expected results unnecessarily. Tests are not passing, but now the expected failure is the same than in all other browsers.
It seems webkit-patch doesn't know how to comment and close bugs with the new version of bugzilla. Committed r214243: <http://trac.webkit.org/changeset/214243>