This is arguable, but if a test is expected to be MISSING, I don't think it should write out the actual results file the way we do for new tests. As it is right now, I end up with junk in my LayoutTests directory every time I run the tests because some tests are listed as MISSING.
the only thing that makes me waffle a bit on this is when we have bugs that cause dumpAsText() to not be called and we report MISSING because we don't find a matching png ... the results are a pretty good clue that something's wrong here.
but we could probably address that differently ... see https://bugs.webkit.org/show_bug.cgi?id=63104 .
Not sure how that's related. This would only affect files listed in TestExpectations as MISSING. The main use-case for writing out the results is for new tests that people write, which obviously wouldn't have a MISSING entry in TestExpectations.
oh, I see. expected missing vs. unexpected missing ...
Right.
We'll be getting rid of MISSING soon in the great layout test refactor. So, this doesn't seem worth fixing. I looked into it and it requires a good deal of plumbing and/or hackery.