On Thu, Sep 20, 2012 at 04:30:55PM -0700, Ethan Jackson wrote: > In ovsdb_jsonrpc_session_got_request() we scope creating of a monitor > to a particular database, but don't scope deletions of monitors in the > same way. Is it possible that two monitors of the same name would be > added to two different databases, and then a cancel could destroy the > wrong one?
I think you've misread some code here. A monitor applies to a single database, but the scope of the monitor-id (the name) is a JSON-RPC session. I think that both the create and the cancel code implements this scope the same way. Can you take another look? > Trailing whitespace in the prototype of parse_options() in ovsdb-server.c. Thanks, I fixed this up. > In update_remote_status() txn is used uninitialized in the call to > update_remote_rows(). Ouch. That's a nasty bug. I see that we didn't have any tests that hit this code. I added a test, but it required a couple of additional changes, so I'll send out a new series. > Finally, it seems a bit awkward to me that we have to NULL terminate > the dbs list just so the unixctl callback doesn't need its length. > I'd slightly prefer to simply make dbs and n_dbs global. That said, > it doesn't matter that much. I think I'll just leave this if that's OK. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
