Timing problem in DROP TABLESPACE?

I've run regression tests many, many times recently, usually in the form
of a looping installcheck, rather than just make check.

On a recent test the last command of the last test has failed:
DROP TABLESPACE testspace;
ERROR:  tablespace "testspace" is not empty

The directory was confirmed as non-empty, though when running the
command again it worked successfully.

However, I've run it many times and it hasn't happened at all, so it's
sporadic and fairly rare. When I isolate just that test case in a tight
loop, I've seen no problems.

Now I'm not sure what causes that and although I am working on patches I
don't think I'm causing it myself. This error happens on the *master*,
not on the standby server.

Is there a timing problem? I see that we attempt to do a checkpoint to
clean up deleted files. Is that working? Is it ignoring certain
messages?

-- Simon Riggs           www.2ndQuadrant.comPostgreSQL Training, Services and Support