Travis posted <address@hidden>, excerpted
below, on Sat, 14 May 2005 12:58:13 -0700:
Every once in awhile PAN crashes and when I start it back up all
the previously "read" articles come back as "unread".
PAN 0.14.2 is running on WinXP SP2.
Is this normal? If not how do I fix it?
It's normal, for current stable and beta versions anyway, altho a
lot is changing in CVS, and therefore for the next versions of PAN.
What's going on with non-CVS, causing the behavior you see, is this.
While in a group, for performance reasons, PAN doesn't write to
disk the status change every time you read a message. It keeps
that info in memory, and normally only flushes it when you change
groups, causing PAN to flush all the changed info on the current
group, and read in all the overview info for the newly loaded group.
Thus, an effective work-around, if PAN isn't stable for you and you
don't like losing track of your read messages when it crashes and
you restart it, is to periodically switch groups (pick one with few
or no messages and it'll load faster), then switch back to the
group you were in. Pick your period. The longer you wait, the
more unsaved read-message data PAN accumulates. So, any time you
come to a convenient stopping place, after you've finished reading
a long thread, for instance, or just marked a whole bunch of stuff
read and don't want to lose that status change, quickly switch to a
different group then reload the one you were working in. Anything
since the last group reload should now be safely saved to disk. If
it isn't, it's your choice of file system or OS cache flushing
rules that's to blame, or your hard drive not flushing its cache,
not PAN's problem.
I /think/, but haven't actually fully verified, that you can reload
the group you are in (and thus flush current status to disk),
without actually switching to another group and thus losing your
place, by simply double-clicking your current group. That /seems/
to reload it here, but I haven't crashed enough while I was
actually tracking doing that vs. switching out and then back, to
fully verify that it saves the full status same as fully switching
out and then back does. So, either experiment yourself, or switch
to an entirely different group and then back, when you want to be
/sure/ it's saved.