Try to smooth out the edge cases for cacheReq not ever updating data.

 Main problems are:

1. User doesn't know cache is N months old.

2. User uses --enablerepo=updates-testing and that's N months older than
current fedora+updates.

cli.py

@@ -394,6 +394,43 @@ def parseCommands(self):
  cacheReq = 'write'
  if hasattr(cmd, 'cacheRequirement'):
  cacheReq = cmd.cacheRequirement(self, self.basecmd, self.extcmds)
+
+ # The main thing we want to do here is that if the user has done a
+ # "yum makecache fast" or has yum-cron running or something, then try
+ # not to update the repo. caches ... thus. not turning 0.5s ops. into
+ # 100x longer ops.
+ # However if the repos. are not in sync. that's probably not going to
+ # work well (Eg. user enables updates-testing). Also give a warning if
+ # they are _really_ old.
+ ts_min = None
+ ts_max = None
+ for repo in self.repos.sort():
+ if not os.path.exists(repo.metadata_cookie):
+ ts_min = None
+ break
+
+ rts = os.stat(repo.metadata_cookie).st_mtime
+ if not ts_min:
+ ts_min = rts
+ ts_max = rts
+ elif rts > ts_max:
+ ts_max = rts
+ elif rts < ts_min:
+ ts_min = rts
+
+ if ts_min:
+ # If caches are within 5 days of each other, they are ok to work
+ # together (lol, random numbers)...
+ if (ts_max - ts_min) > (60 * 60 * 24 * 5):
+ ts_min = None
+ elif ts_max > time.time():
+ ts_min = None
+
+ if not ts_min:
+ cacheReq = 'write'
+ elif (time.time() - ts_max) > (60 * 60 * 24 * 14):
+ self.logger.warning(_("Repodata is over 2 weeks old. Install yum-cron? Or run: yum makecache fast"))
+
  for repo in self.repos.sort():
  repo._metadata_cache_req = cacheReq