I have investigated the issue and it looks to me the problem is under QNetwork layer. The test case will send several http requests in a row. The first one to "setup.php" contains the login/password info and it went through. The second request to "iframe.php" does not contain user/password in its url and the QNetwork layer is supposed to fetch the credential from its cache. However, I didn't see the credential was saved to cache before this. I think after the successful return of the 1st request, the credential should be saved.

I reported a bug in Qt: http://bugreports.qt.nokia.com/browse/QTBUG-9996

the bug is fixed in Qt4.7. Once the QtWebKit buildbot starts using Qt4.7, we can close this bug.

Hey, Can i then close the associated bug in the Qt bugtracker? http://bugreports.qt.nokia.com/browse/QTBUG-9996 Thanks

Created attachment 77147 [details] Patch to unskip auth.html from Skipped. Appcache layout test case auth.html can be removed from the Skipped list. Test case passes since switching to Qt4.7 in the buildbot.

The commit-queue encountered the following flaky tests while processing attachment 77147 [details]: http/tests/local/stylesheet-and-script-load-order-media-print.html bug 51470 (author: koivisto@iki.fi) The commit-queue is continuing to process your patch.

Comment on attachment 77147 [details] Patch to unskip auth.html from Skipped. Clearing flags on attachment: 77147 Committed r74481: <http://trac.webkit.org/changeset/74481>

All reviewed patches have been landed. Closing bug.

It seems to me that running http/tests/appcache/auth.html makes the http/tests/misc/401-alternative-content.php flaky.

http://trac.webkit.org/changeset/74481 might have broken GTK Linux 64-bit Debug The following tests are not passing: http/tests/local/stylesheet-and-script-load-order-media-print.html

The patch was reverted by http://trac.webkit.org/changeset/74486 as it causes regression. A re-work is going on.

What was the regression? The stylesheet test failure? If so, please dup bug 51470 against this one?

*** Bug 51829 has been marked as a duplicate of this bug. ***

(In reply to comment #13) > *** Bug 51829 has been marked as a duplicate of this bug. *** After http://trac.webkit.org/changeset/74896 http/tests/misc/401-alternative-content.php fails always if I run all of the tests, but pass if I run only this test. So it must be a Qt-DRT sideeffect bug, and not flakyness as Laszlo said. pretty diff: http://build.webkit.org/results/Qt%20Linux%20Release/r74900%20%2825916%29/http/tests/misc/401-alternative-content-pretty-diff.html

Created attachment 80449 [details] fix patch

Comment on attachment 80449 [details] fix patch View in context: https://bugs.webkit.org/attachment.cgi?id=80449&action=review r=me, with one adjustment: > Tools/DumpRenderTree/qt/DumpRenderTreeQt.cpp:560 > + if (isHttpTest(url)) { The helper method is a bit misleading, I think we should simply inline the check here, and also make it more specific, i.e: if (url.scheme() == "http" || url.scheme() == "https")

Laszlo suggested to do network cleanup for all tests, including non-http. This clean up the code. I ran part of the tests so far and no failures and no performance overhead observed. Andreas, do you prefer this change?

(In reply to comment #17) > Laszlo suggested to do network cleanup for all tests, including non-http. This clean up the code. I ran part of the tests so far and no failures and no performance overhead observed. Andreas, do you prefer this change? If it has no tangible performance overhead, sure.

Created attachment 80456 patch based on Kling's comments.

Created attachment 80457 [details] fix patch 2

Comment on attachment 80457 [details] fix patch 2 Clearing flags on attachment: 80457 Committed r76958: <http://trac.webkit.org/changeset/76958>

All reviewed patches have been landed. Closing bug.