This layout test: http://trac.webkit.org/browser/trunk/LayoutTests/fast/events/script-tests/onunload-back-to-page-cache.js?rev=64939 Sets window.name to 'navigated' before navigating to a page with an onunload handler. Mac (and other platform) DRT prints this message: main frame "navigated" - has 1 onunload handler(s) While GTK DRT prints this message: main frame - has 1 onunload handler(s) Presumably getFrameNameSuitableForTestResult in WebKitTools/DumpRenderTree/gtk/DumpRenderTree.cpp should be brought closer to _drt_descriptionSuitableForTestResult in WebKitTools/DumpRenderTree/mac/FrameLoadDelegate.mm
I don't see this problem locally. When I revert the change you made in r64953, I get a failure, because "navigated" is listed in the output. Can you recall if this was failing on all bots or just some of them? Perhaps it's another timing issue.
I only happened to check the GTK Linux 64-bit Debug bot, but it seemed to be failing consistently. If you revert r64953 (for both the test and the expected output) and get a failure, doesn't that mean that you can reproduce this? "navigated" should be in the output in that case (and did appear to show up in all the other bots).
This test doesn't fail anymore on any platform, it seems though all the ports are now outputting the onunload handler report without the frame name. http://trac.webkit.org/browser/trunk/LayoutTests/fast/events/onunload-back-to-page-cache-expected.txt