The http/tests/appcache/different-https-origin-resource-main.html layout test is failing on GTK+. It was expected to crash but it is failing after r[r184624-r184632]. The diff is the following: --- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/http/tests/appcache/different-https-origin-resource-main-expected.txt +++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/http/tests/appcache/different-https-origin-resource-main-actual.txt @@ -1,4 +1,4 @@ Test that a resource from a different https origin is not cached. -PASS +ERROR If I try to print the error information I get an empty object.
Created attachment 270428 [details] GDB Backtrace for the crash of http/tests/appcache/different-https-origin-resource-main.html on a GTKDebug build of r195954
I have observed the following: If the full test suite is run, this tests usually fails. But if this tests is run alone by specifying its name on the command line it crashes. I'm attaching the GDB backtrace for when it crashes.
Created attachment 270489 [details] Patch There's a use after free when handling tls errors.
Comment on attachment 270489 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=270489&action=review Thanks Carlos and Carlos. Yet another concrete result to show from our gardening. > Source/WebCore/ChangeLog:3 > + [GTK] Layout Test http/tests/appcache/different-https-origin-resource-main.html is failing Also remove the failure expectation?
(In reply to comment #4) > > Also remove the failure expectation? It continues to fail. This patch fixed the crash, but not the test. So I guess that removing the Crash from the expectation and leaving the Failure is the right thing to do.
Committed r196061: <http://trac.webkit.org/changeset/196061>
Reopening because the test is still failing and the expectation is still marked against this bug. Glad we got the crash fixed though.
Fixed by r228892.
<rdar://problem/37959838>