nrwt: fix rebaselining for a generic directory
Created attachment 82865 [details] Patch
Created attachment 82867 [details] take two
Created attachment 82875 [details] Patch
Okay, this third patch is close to the "right" fix. I put "right" in quotation marks because there are at least two things wrong with using rebaseline-chromium-webkit-tests these days that keep it from being truly safe. The first is that we don't have a full complement of stable bots on the different versions (Win 7, Vista, SL), and so we hack around this in order to get the baselines in the right place. What should happen is that if you want something to go into chromium-win, you should either rebaseline the WIN-7 version of the test, or rebaseline all of the versions of the test and rely on the deduping logic to leave a single copy. The second is that I believe chromium-win-win7 is actually using platform/chromium-win-win7 as the baseline_path(), which is just wrong. I'm not fixing this now because I have another set of patches coming up that will fix this and continue to make the version code cleaner and more consistent across ports, and I'll fix it as part of that. We have a similar, but more debatable, issue w/ snow leopard (debatable because SL should either be using chromium-mac-snowleopard or chromium-mac depending on whether you think "lion" is a real port or not yet.
Created attachment 83010 [details] clarify behavior of baseline_path for chromium-mac-snowleopard, chromium-win-win7
Committed r79035: <http://trac.webkit.org/changeset/79035>