[Pan-users] Connection pool oddity

[Jim Reiss]

I apologize if this has been reported before...I did a quick search and
also don't recall seeing anything about it.

Anyway, with the new beta I'm trying out the connection pool feature for
the first time, and I noticed an oddity with how it deals with reverting
to a previous task list after I kill pan.  My primary news server is
configured with 4 connections.  When pan starts up and I tell it to
revert it first starts trying to download three different tasks (which
mirrors the old behavior where it was reserving one connection).  It
soon realizes that this isn't the way it's supposed to work now and it
stops downloading two of the three (looks like it downloads one segment
of each and then the connections reallocate).  So I end up with one task
running (with multiple connections) and two tasks that say they're
queued with a couple of percent already downloaded.

The first time I saw this pan actually killed itself with a segmentation
fault, as happens to me now and then when I'm using "find by subject".
If I clear the selection criterion and then select another group it
always crashes if the new group loads completely before it can apply the
new filter.  Just mentioning it because it's very similar to a problem
that was supposedly fixed some time back, but I keep seeing it.  Pan
loads the new group then finishes applying the filter to the original
group, switching back to displaying those headers.  If I'm quick I can
prevent a crash by quickly selecting a group with few headers (loads
quickly), but usually I end up having the segfault.  I always kick
myself for not working around a problem I know about, but sometimes you
just do what seems intuitive ("I don't want this filter in the new
group, so clear it, press enter, and switch").