Nevertheless, it actually wouldn't hurt to add a couple of interactive tests, too. Having non-interactive tests is nice, but they don't help to notice visual bugs.
That's true, I just didn't know how to do that. I'll probably add a few once this implementation is finished.
Yes, you should send your patches in smaller chunks. Also, you shouldn't rely on the assumption that all of your patches are applied at once. The patchset should still make sense, even if each commit is applied separately. If it doesn't, that would mean you have to reorder or merge them. As long as you mark failing tests with todo_wine it is not a problem to have a temporary state with "less features" than the original implementation though.
Sorry, I still don't understand what's wrong. The patches depend on each other, so just applying patch 9 would result in an error. But only applying 1-5 should pose no problem. Except for missing functionality of course, but the tests should be fine. But I have to send these 12 patches as single patchset, since it's supposed to replace the old taskdialog-hack and I don't want to introduce a regression. Only with all those patches the functionality is the same.
Regards, Fabian Maurer