We had a problem with two migrations not working nicely on sqlite, because the second one wanted to do an alter table and sqlite doesn't like to do that in a bigger transaction. Long story: Every migration should say "I'm done" and commit. This happened for the existing migrations.

So. In my opinion, the migration framework should run a rollback after each migration "just in case" to clean up any cruft left over from a (possibly broken) migration.

- migration_func(self.session)
+ try:
+   migration_func(self.session)
+ finally:
+   self.session.rollback()

I have NOT TESTED this patch. This is just to show the idea. I want some feedback on it, and possibly someone to test it.

this error happened with two different videos in two different formats (avi and webm)

[2012-06-22 22:22:36,020: INFO/MainProcess] Got task from broker: mediagoblin.processing.task.ProcessMedia[7e73cf15-8594-41c7-93f4-163bcbb662e0]
[2012-06-22 22:23:07,291: INFO/PoolWorker-2] Initializing VideoTranscoder...
[2012-06-22 22:23:29,388: INFO/PoolWorker-2] Discovering...
[2012-06-22 22:23:31,666: INFO/PoolWorker-2] Done
[2012-06-22 22:23:33,999: WARNING/PoolWorker-2] Traceback (most recent call last):
[2012-06-22 22:23:34,229: WARNING/PoolWorker-2] File "/srv/http/mediagoblin/mediagoblin/mediagoblin/media_types/video/transcoders.py", line 441, in __discovered
[2012-06-22 22:23:34,704: WARNING/PoolWorker-2] raise Exception('Could not discover {0}'.format(self.source_path))
[2012-06-22 22:23:34,787: WARNING/PoolWorker-2] Exception
[2012-06-22 22:23:34,804: WARNING/PoolWorker-2] :
[2012-06-22 22:23:34,804: WARNING/PoolWorker-2] Could not discover /srv/http/mediagoblin/mediagoblin/user_dev/media/queue/media_entries/7e73cf15-8594-41c7-93f4-163bcbb662e0/Gurren_Lagann_HD_Creditless_Opening_1.webm
[2012-06-22 22:23:34,880: INFO/PoolWorker-2] Terminating MainLoop
[2012-06-22 22:23:34,880: DEBUG/PoolWorker-2] Saving medium...
[2012-06-22 22:23:35,316: DEBUG/PoolWorker-2] Saved medium
[2012-06-22 22:23:51,120: WARNING/PoolWorker-2] No idea what happened here, but it failed: AttributeError("VideoTranscoder instance has no attribute 'dst_data'",)
[2012-06-22 22:23:55,970: ERROR/MainProcess] Task mediagoblin.processing.task.ProcessMedia[7e73cf15-8594-41c7-93f4-163bcbb662e0] raised exception: AttributeError("VideoTranscoder instance has no attribute 'dst_data'",)
Traceback (most recent call last):
  File "/srv/http/mediagoblin/mediagoblin/lib/python2.7/site-packages/celery-2.5.3-py2.7.egg/celery/execute/trace.py", line 181, in trace_task
    R = retval = fun(*args, **kwargs)
  File "/srv/http/mediagoblin/mediagoblin/mediagoblin/processing/task.py", line 52, in run
    manager['processor'](entry)
  File "/srv/http/mediagoblin/mediagoblin/mediagoblin/media_types/video/processing.py", line 88, in process_video
    width=transcoder.dst_data.videowidth,
AttributeError: VideoTranscoder instance has no attribute 'dst_data'

The current media processing panel we have is a demo more than anything else, and I'm not really sure it works. But users want something like this: something that's an indicator that their things are processing and to see whether things processed or failed. Administrators also should be able to see detailed errors.

This will be mostly a programming task, but we'll need some careful graphic design too to make this look nice.