They run and always say they pass. For example, for fast/writing-mode/borders.html, modify it and see that it still passes even though borders-expected.html now renders differently. They seem to run correctly on the chromium mac build at least.
This reminds me of this bug. https://bugs.webkit.org/show_bug.cgi?id=59188 I guess a regression might happen. Let me take a look later. I can not access to a Mac environment for a week.
We should make it so that running a reftest without -p either does the pixel comparison or prints an error.
Relevant webkit-dev discussion: https://lists.webkit.org/pipermail/webkit-dev/2011-November/018422.html So far it sounds like people are not opposed to just enabling reftests by default. I guess we'll eventually need to provide a --no-reftest or something to disable running reftests since --no-pixel doesn't do it. I'm OK with just enabling it by default for now and seeing if people ever actually need to disable them.
Fixing bug 60605 will effectively resolve thisl *** This bug has been marked as a duplicate of bug 60605 ***