Remove base-* directories from gm expected/actual paths; just use platform names

buildbot/slave/skia_slave_scripts/build_step.py

Index: buildbot/slave/skia_slave_scripts/build_step.py
===================================================================
--- buildbot/slave/skia_slave_scripts/build_step.py	(revision 10698)
+++ buildbot/slave/skia_slave_scripts/build_step.py	(working copy)
@@ -169,9 +169,9 @@
     self._gm_actual_basedir = os.path.join(os.pardir, os.pardir, 'gm', 'actual')
     self._gm_merge_basedir = os.path.join(os.pardir, os.pardir, 'gm', 'merge')
     self._gm_expected_dir = os.path.join('expectations', 'gm',
-                                         self._gm_image_subdir)
+                                         self._builder_name)

[epoger, 2013/08/13 20:33:57]

Patchset 3 updates the buildbot slave scripts to use builder_name instead of
gm_image_subdir to construct paths when reading/writing gm results.

(The skp code, like in skia_slave_scripts/playback_dirs.py , continues to use
gm_image_subdir to construct its paths.  That will change sometime, hopefully
soon.)

[borenet, 2013/08/14 12:57:49]

I guess the SKP code causes a wrinkle, but I would prefer that we continue to
use self._gm_image_subdir instead of self._builder_name.

[epoger, 2013/08/16 20:16:25]

So, once we have made both the GM and SKP expectations "per-builder", are you
proposing that we should store the builder name as both self._builder_name and
self._gm_image_subdir ?  That seems redundant to me.

I think it would be simpler, at that point, for every slave script which
reads/writes expectations to assume that builder_name is the proper key.

     self._gm_actual_dir = os.path.join(self._gm_actual_basedir,
-                                       self._gm_image_subdir)
+                                       self._builder_name)
     self._gm_actual_svn_baseurl = '%s/%s' % (args['autogen_svn_baseurl'],
                                              'gm-actual')
     self._resource_dir = 'resources'