Bug 108416 – consider adding a platform/mac-wk1 directory

• Status: RESOLVED
• Resolution: WONTFIX
• Component: Tools / Tests

[Dirk Pranke]

for WK1 Apple Mac-specific baselines and exceptions. 

This appears to come up from time to time (e.g., bug 106187 just now), where there is no good solution to indicate that a test will fail on WK1 and Lion WK2 but should work on newer versions of WK2.

Other ports (Gtk, Qt, Efl) have both -wk1 and -wk2 directories, but apple mac only uses -wk2. Adding a -wk1 directory would solve this problem.

However, the issue is complicated slightly by the fact that -- as things are currently implemented -- the easiest patch to do this would change the default location for new baselines from platform/mac-$version to platform/mac-wk1. Also, when generating new baselines with -2, we have that problem today (baselines go into mac-wk2 instead of platform/mac-$version).

It seems to me that using those directories by default is probably wrong, and I'm actually surprised this hasn't come up yet, but maybe baselines are more often generated w/ the -wk1 builds?

I could change the code to never use the -wk1 or -wk2 variants for baselines by default (and you'd have to manually specify a directory to use it), or I could change the code to leave -wk2 alone and just never use -wk1.

Or we could go whole-hog and create mac-$version-{wk1,wk2} dirs and use those as appropriate and rely on the baseline optimizer to de-dup things.

(Note that Gtk, Qt, Efl don't have this problem since they don't generally have version-specific dirs).

I have no real skin in this game, so I will implement whatever the apple folks want here.

[Dirk Pranke]

Note that there is also the option to add "WK1" and "WK2" modifiers in the TestExpectations file, but as per the most recent thread on this the consensus seemed to be against this.

https://lists.webkit.org/pipermail/webkit-dev/2012-November/022825.html

[Dirk Pranke]

Created attachment 185656 [details]
Patch

[Dirk Pranke]

The failing test in question is:

http://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/layout_tests/port/mac_unittest.py#L127

(another test fails w/ the same problem, but this one illustrates the problem sufficiently).

[Eric Seidel (no email)]

This is definitely a patch for review by one of the Mac WK1 maintainers.

[Dirk Pranke]

smfr, jberlin, rniwa ... any thoughts? Failing that, I'll just abandon this.