bug in mutable publish that could cause an IndexError when a writer is removed in Publish._connection_problem

[davidsarah]

"Traceback (most recent call last):
Failure: allmydata.mutable.common.NotEnoughServersError: (\"Publish ran out of good servers, last failure was:
[Failure instance: Traceback: <type 'exceptions.IndexError'>: list index out of range
/home/davidsarah/cloud-branch/support/lib/python2.6/site-packages/Twisted-12.0.0-py2.6-linux-i686.egg/twisted/internet/base.py:800:runUntilCurrent
/home/davidsarah/cloud-branch/support/lib/python2.6/site-packages/foolscap-0.6.3-py2.6.egg/foolscap/eventual.py:26:_turn
/home/davidsarah/cloud-branch/support/lib/python2.6/site-packages/Twisted-12.0.0-py2.6-linux-i686.egg/twisted/internet/defer.py:368:callback
/home/davidsarah/cloud-branch/support/lib/python2.6/site-packages/Twisted-12.0.0-py2.6-linux-i686.egg/twisted/internet/defer.py:464:_startRunCallbacks
--- <exception caught here> ---\\n/home/davidsarah/cloud-branch/support/lib/python2.6/site-packages/Twisted-12.0.0-py2.6-linux-i686.egg/twisted/internet/defer.py:551:_runCallbacks
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:634:_push
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:651:push_segment
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:637:_push
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:773:push_everything_else
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:878:finish_publishing
/home/davidsarah/cloud-branch/src/allmydata/mutable/publish.py:886:_record_verinfo

I can reproduce this, at least on the cloud-branch, when I do a tahoe put --mutable shortly after the gateway has started.

[davidsarah]

zooko suggests this fix, which I reviewed and approved of:

--- old-dw/src/allmydata/mutable/publish.py     2012-05-22 21:48:48.764939441 -0600
+++ new-dw/src/allmydata/mutable/publish.py     2012-05-22 21:48:50.641598788 -0600
@@ -253,6 +253,10 @@
         # updating, we ignore damaged and missing shares -- callers must
         # do a repair to repair and recreate these.
         self.goal = set(self._servermap.get_known_shares())
+
+        # k: shnum, v: [ instance of IMutableSlotWriter ]
+        # The value is required to always be non-empty if the item is present
+        # in the dict at all.
         self.writers = {}
 
         # SDMF files are updated differently.
@@ -891,8 +895,10 @@
         """
         self.log("found problem: %s" % str(f))
         self._last_failure = f
-        self.writers[writer.shnum].remove(writer)
-
+        writers = self.writers[writer.shnum]
+        writers.remove(writer)
+        if len(writers) == 0:
+            del self.writers[writer.shnum]
 
     def log_goal(self, goal, message=""):
         logmsg = [message]