Merge lp:~smoser/cloud-init/filesem-check-canon into lp:~cloud-init-dev/cloud-init/trunk

[Scott Moser]

[Joshua Harlow (harlowja)]

+1 Seems good, maybe in a future version we can move the non-canon 'files' to there rightful 'canon' name if we encounter them, thus avoiding repeated checks in the future.

[Joshua Harlow (harlowja)]

After our discussion:

https://code.launchpad.net/~harlowja/cloud-init/migrator-module/+merge/133409

Maybe we want to go forward with that instead?

[Scott Moser (smoser)]

This was marked 'rejected' as we used the migrator module instead.

Unmerged revisions

701. By Scott Moser on 2012-11-07

check for canonicalized name in file semaphore checking

This works towards a fix for 1075980 by at least looking for filename
markers for a "name" with or without '-' replaced with '_'.

Ie, if 'name' is "foo-bar", this will check for both existance of
a file marker 'foo-bar' and 'foo_bar'.

=== modified file 'cloudinit/helpers.py'
--- cloudinit/helpers.py	2012-10-28 02:25:48 +0000
+++ cloudinit/helpers.py	2012-11-07 16:56:21 +0000
@@ -107,7 +107,7 @@
         # This is a race condition since nothing atomic is happening
         # here, but this should be ok due to the nature of when
         # and where cloud-init runs... (file writing is not a lock...)
-        sem_file = self._get_path(name, freq)
+        sem_file = self._get_path(_canon_sem_name(name), freq)
         contents = "%s: %s\n" % (os.getpid(), time())
         try:
             util.write_file(sem_file, contents)
@@ -119,11 +119,15 @@
     def has_run(self, name, freq):
         if not freq or freq == PER_ALWAYS:
             return False
-        sem_file = self._get_path(name, freq)
+
+        # check for both the name, and canonicalized name
+        cand_list = set([self._get_path(n, freq)
+                        for n in [name, _canon_sem_name(name)]])
         # This isn't really a good atomic check
         # but it suffices for where and when cloudinit runs
-        if os.path.exists(sem_file):
-            return True
+        for cand in cand_list:
+            if os.path.exists(cand):
+                return True
         return False
 
     def _get_path(self, name, freq):
@@ -426,3 +430,7 @@
             if header:
                 contents = "\n".join([header, contents])
         return contents
+
+
+def _canon_sem_name(name):
+    return(name.replace("-", "_"))