[Webkit-unassigned] [Bug 50098] New webkit-patch rebaseline2 command.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=50098





--- Comment #17 from Dirk Pranke <dpranke at chromium.org>  2010-12-02 17:52:08 PST ---
(In reply to comment #14)
> I was wondering if it was possible / a good idea to change our test fallback order so that it does make a tree? 

Yes, it is theoretically possible. I have no idea how much work it would be. I do think it would be a good idea to evaluate it and see, because the current logic causes no end of confusion.

Not as part of this patch, though :)

> 
> I'm not sure it's easy to use baseline_search_path() instead of the current code - if I just use one port's baseline_search_path() then it won't be broad enough to be useful[1], and if combine all the baseline_search_path()s then it doesn't form a tree. I'll investigate making my deduping algorithm work for DAGs, and then I can generate the DAG from the baseline_search_paths(), and then as a next step replace the baseline_search_path() implementations with a single one that walks the DAG to calculate its fallback path.
> 
> [1] eg: I'm on snowleopard now, so my baseline_search_path() is [mac-snowleopard, mac] which should never cause deduping because we don't know anything about mac-leopard or the other paths that fallback onto mac.

Yes, Victor and I had a lot of discussions about how to do this when he was implementing/maintaining rebaseline-chromium-webkit-tests. You end up carrying around multiple Port objects - the port you're actually running on (to be able to diff images, etc.), the Port you're manipulating results for, and I think we even needed another one for some reason. I suggest you look carefully at that code.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.