On Fri, 2010-07-16 at 10:11 -0400, Charles Gregory wrote:
> Now if I have to GUESS on insufficient data, I would suspect that the
> 'port' of spamd to Windows(?) does not properly tidy up its children when
> finished. The fact that it crashes certainly points in this direction.
> May I presume that you did a 'full' memory test?
>
> To verify this situation, try running the same test as before, but leave a
> one minute gap between each run/test (and with no other spamd calls during
> that time interval!) so that we can see what happens when the spamd
> children have time to properly terminate.
>
You might also do a pair of test runs with the same set of test data and
the options shown:
- one with --max-children=1 which should force sequential scans using
the same child. This will pick up any cruft being left in the child
process by the previous message.
- one with --max-children=1 and --max-conn-per-child=1 which should
force a newly spawned child to be used for every message.
Any differences between the two runs would point to left-over cruft
being the problem.
Martin
