supervisor/TODO.txt

- Web interface:

   - present status in addition to state (e.g. spawnerr, etc, same as
     supervisorctl)

   - provide "restart all" functionality

- Documentation:

   - In nodaemon mode, child processes will be sent SIGINT when Ctrl-C
     is pressed.

   - Error conditions when X-bit not set on command file.

- Unit test the http_client package.

- Unit tests for meld classes and ui server.

- Create named process log files after we setuid.

- Change xmlrpc_user/xmlrpc_password to http_X.

- Usage messages when we invoke supervisord/supervisorctl -h

- Tail main log.

- Figure out how to test the actual supervisord class.

- Test log rotation.

- Command-line arg tests.

- supervisorctl one-shot tests.

- packaging.

- test under py 2.3 to make setup.py assertions true

- supervisorctl command history

- test startup as root

- stop logging all RPC requests in info mode when we ship

- reopen child logs when we get a logreopen signal

- Collector issue 75:

  If the pid file isn't writable, it dies silently. When I ran it with
  --nodaemon this became clear, but nothing about the error goes into
  the log file or console when daemonizing.

- Collector issue 76:

  Adding -c to the shell script also drove me nuts for a while, since
  I got an error about two -c's. Maybe instead it could set an
  environmental variable in the shell script and read that for the
  default?

- Collector issues 78:

  The logs fill up with "INFO waitpid error!" (on FreeBSD) when you
  don't have any processes started. This happens when you don't have
  auto-start on, or just when the last process has temporarily died.

- Collector issue 84:

  I'd like to make Paste detect when it is running under Supervisor,
  mostly so it would know that it could safely die to be
  restarted. Otherwise it starts its own monitoring process to handle
  restarts.  I think always setting an environmental variable (e.g.,
  "SUPERVISOR_ENABLED") would be sufficient.

 - Collector issue 152:

   I'd like supervisor to look after a number of postgres instances
   (well, initially only 1, but I will need it to manage 3). The first
   difficulty I had was that the postgres start-up program (pg_ctl)
   takes an options param that it passes directly to the postmaster
   program it controls, which can contain several options that only
   the postmaster understands and it expects these to be in quotes so
   they are treated as a group, e.g pg_ctl -o "-p 4321 -h hostname -d
   datadir" but supervisor doesn't appear to preserve the quotes. In
   my case I only need the one option so I used -o "-p4321", as a
   workaround.