[Webkit-unassigned] [Bug 58122] chromium rebaseline script did the wrong thing with snowleopard expectations

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=58122





--- Comment #4 from Dirk Pranke <dpranke at chromium.org>  2011-04-12 14:11:28 PST ---
(In reply to comment #3)
> The way I was picturing this working is the following:
> When rebaselining any platform's results, we find the first result in it't it's baseline search path and copy it to that platform's result. Then we grab the results off the bot and dedupe. It's a good deal more work on the filesystem, but I don't think the filesystem work is significant in terms of how long this script takes to run.
> 
> So, in this case, to start with, there's a result in chromium-mac and we want to rebaseline snowleopard. So we do the following:
> 1. Copy the result from chromium-mac to chromium-mac-leopard and chromium-mac-snowleopard.

I'm not sure I follow this algorithm. Partially, I think this is because your example doesn't match the previous paragraph. If we were rebaselining just SL, why would be touch chromium-mac-leopard (since that's not in the path)?

But, I'll assume that's a typo and try to interpret the first paragraph. 

The way the code works now, it processes a port at a time, depth first, so SL is processed, then Leopard, then Win7, then Vista, etc. 

So, first we run SL, and SL updates chromium-mac with the new baseline. Then we run Leopard, and at that point have lost access to the original chromium-mac baseline (which was really a Leopard baseline).

To fix this, we'd have to do multiple passes down the port list first to copy the baselines into the version-specific dir, and then update all of the baselines, and then de-dup in a third pass (might be able to copy this into the second pass).

I'm a bit worried that this might be a fair amount of surgery on the script and might destabilize other things. Given that this will likely only come up when we introduce new versions of operating systems, I'm not sure if it's worth the effort. Thoughts?

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