[Pan-users] Re: quick notes

[Duncan]

K. Haley posted on Sun, 10 Apr 2011 01:55:50 -0600 as excerpted:

> 1. The cross posting problem:
>   Pan keeps track of the highest xover article number in the xov file.
> It then retrieves headers from this value up to the servers current
> high.  This value is only changed when retrieving headers.  The only
> times headers get removed are when you delete them and when they expire.
>  You could grep the group file for one of the missing articles
> message-id's.  You could also try set pan to show matched articles,
> I've had occasional trouble with the show matched & threads setting.
> It's easy to spot when I see an unread count and no articles in the
> tree.

The interesting thing here is that for affected groups, it's almost as if 
pan doesn't even try.  Given your description, it's as if it it's using a 
highwater xover from some other, much more active group.

WAIT!!!

I just tried getting all headers for an affected group, and DESPITE the 
fact that it's a gmane group, gmane.* hierarchy, not carried by anything 
BUT gmane, and DESPITE the fact that I have gmane set to "never expire", 
the log said "expiring 64 headers for <group>, and the group is again 
entirely empty.

Re-fetching all headers doesn't seem to do anything at all. ??

Maybe it's the expiration going haywire.  Check to see if there's a low-
use code-path where a "0" expiration is taken as literally zero days 
(thus, immediately), instead of don't expire at all.

Actually, that might explain the cross-posting link as well, since there 
may be some logic that treats it differently.  (Consider the case of a 
cross-posted message retrieved from a short-expiration server, whether one 
of the groups posted to only appears on a longer-expiration server.  
Having the message disappear from the group only on the long-expiration 
server at short-expiration-server time is arguably not a good thing, tho I 
could see it argued either way.  If Charles had some code intended to deal 
with that sort of thing, that only applied to cross-posts...)

That could explain why it hits so few people, too.  It's likely that 
relatively few have don't-expire-at-all set on their servers.  Combined 
with an apparent cross-posting trigger, it'd only hit certain groups for 
only those users with no-expiration set on at least one server.  That's 
likely a fairly small cross-section.

> 2. Mainly a note to myself to set up a timer to automatically save the
> newsrc files after articles have been marked.

=:^)

> 3. I haven't looked at any of the signature code.  As far as I know my
> work on text wrapping shouldn't affect it.

I believe you're right.  But it's possible that some of the patches you 
applied for regular body-text need to be applied to the the sig-handing 
code as well -- that's what I was suggesting, not that your patches had 
screwed it up.

Thinking about it tho, I expect Charles' intent was to not wrap the sig at 
all, adding it as-is, and the bug is that some of the wrapping code is 
applying to it anyway, when it shouldn't.  Because that wasn't intended, 
it's doing something weird... duplicating it!

Since Joe Zeff says it happened to him some time back, presumably 0.133 or 
even earlier, the bug must be in Charles' code, not yours.

> I've been really busy lately so I haven't had time to work on pan.

"Real Live" (R) has a way of doing that to people. =:^s


Meanwhile, if that x-post bug DOES end up being expiration, and can be 
traced and fixed now, I'll be a happy camper indeed, because that bug has 
been bugging me since that year plus of heavy activity that started with 
the C++ rewrite and 0.90 and ended at 0.130 or there abouts.  I'm guessing 
0.120 or 0.125 or so.  With that severe a bug, I didn't really think pan 
was ready to go stable again, much less 1.0, but what could I do, as I 
hadn't a clue on how to narrow it down further and I was the only one who 
seemed to have it?

So if that's fixed, I'll be one HAPPY camper indeed, especially since that 
was sort of my personal blocking bug for 1.0, tho I couldn't really push 
it since I couldn't get enough data to even properly file a bug on it!

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman