The bug 505311 landing to switch SeaMonkey to tabbed browsing by default made some docshell tests in mochitest-plain-2 fail: 215 ERROR TEST-UNEXPECTED-FAIL | /tests/docshell/test/navigation/test_not-opener.html | [SimpleTest/SimpleTest.js, window.onerror] An error occurred - win.location is null at http://mochi.test:8888/tests/docshell/test/navigation/NavigationUtils.js:180 237 ERROR TEST-UNEXPECTED-FAIL | /tests/docshell/test/navigation/test_reserved.html | [SimpleTest/SimpleTest.js, window.onerror] An error occurred - win.location is null at http://mochi.test:8888/tests/docshell/test/navigation/NavigationUtils.js:180 250 ERROR TEST-UNEXPECTED-FAIL | /tests/docshell/test/navigation/test_sibling-off-domain.html | [SimpleTest/SimpleTest.js, window.onerror] An error occurred - win.location is null at http://mochi.test:8888/tests/docshell/test/navigation/NavigationUtils.js:180 I finally tracked that down to the SeaMonkey tab restore function, that keeps the browsers around for some time and restores them from bfcache. If I switch that off (we have a pref for that), the tests pass. Apparently we keep those docshells around after tabs are closed but those preserved docshells have a null window.location and that's what causes http://mxr.mozilla.org/comm-central/source/mozilla/docshell/test/navigation/NavigationUtils.js#180 to error out. As we know we can safely ignore those docshells, we can just add a null check there, and that fixes the tests.

This patch fixes those tests here by simply null-checking win.location.

Comment on attachment 460915 [details] [diff] [review] just null-check win.location r=me. Good catch.

Thanks for the fast review, pushed as http://hg.mozilla.org/mozilla-central/f9ce465b7276

I've no idea why you're getting null location. I expect and see about:blank

Neil, when running those tests _after_ some other tests ran and already loaded and closed tabs, I at least ended up with that function getting docshells with null window.location objects, while with our prefs set to open windows, we don't see those and when I turn off our tab undo cache, I also succeed, so it definitely needs to be something with those browsers we store for undo.