rebaseline.py: re-record already succeeding test results, just in case

tools/rebaseline.py

Index: tools/rebaseline.py
===================================================================
--- tools/rebaseline.py	(revision 10120)
+++ tools/rebaseline.py	(working copy)
@@ -216,7 +216,11 @@
         actuals_url = '/'.join([self._actuals_base_url,
                                 subdir, builder, subdir,
                                 self._actuals_filename])
-        sections = [gm_json.JSONKEY_ACTUALRESULTS_FAILED]
+        # In most cases, we won't need to re-record results that are already
+        # succeeding, but including the SUCCEEDED results will allow us to
+        # re-record expectations if they somehow get out of sync.
+        sections = [gm_json.JSONKEY_ACTUALRESULTS_FAILED,
+                    gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED]

[borenet, 2013/07/17 13:21:28]

I'm confused - this enables comparison between the actual and expected checksums
for GMs which supposedly succeeded, so that we can fix them if they somehow
mismatch?  When would this occur?  The case that comes to mind is if the
expectations are updated after the actuals are generated but before this script
is run.  But if we assume that expectations are only ever updated by this
script, can that ever occur?

         if self._add_new:
             sections.append(gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON)
         results_to_update = self._GetActualResults(json_url=actuals_url,