uriloader/exthandler/tests/mochitest/test_handlerApps.xhtml opens a webcal: URL. On the Mac this opens iCal, which grabs the focus and prompts the user to subscribe to some bogus event. This is kind of silly, obnoxious behavior for an automated test. It would be better to use a dummy app.

Back when we were first testing the web-application support, we used this test application: http://handler-test.mozilla.org/webcal?url=%s It's still active. Do we want to use that instead?

The particular test Jason's mentioning is specifically to check that OS-native app launching works (as much as it's possible to check that).

Hope it doesn't matter to the test whether or not iCal is already running, since thanks to 10.7's annoying habit of reopening programs that were running at the time of a restart, I think (based on screenshots from test hangs in mochitest chunks that aren't the one with this test in it) that we are permanently running iCal on every 10.7 slave once they have run this test once. Is there any less awkward way to kill it on our buildbot slaves than to have the http://mxr.mozilla.org/mozilla-central/source/uriloader/exthandler/tests/mochitest/handlerApps.js#115 chunk check for a pref that buildbot will set while starting mochitests?

Filed bug 743594 for ducking the problem by persuading the 10.7 slaves not to restore running apps.

(In reply to Phil Ringnalda (:philor) from comment #4) > Is there any less awkward way to kill it on our buildbot slaves than to have > the > http://mxr.mozilla.org/mozilla-central/source/uriloader/exthandler/tests/ > mochitest/handlerApps.js#115 chunk check for a pref that buildbot will set > while starting mochitests? Ftr, { 115 // if we ever decide that killing iCal is the right thing to do, change 116 // the if statement below from "NOTDarwin" to "Darwin" 117 if (osString == "NOTDarwin") { ... }