Re: nnimap fails to fetch some articles

[Bjorn Solberg <bjorn_ding@hekneby.org>]

Simon Josefsson writes:

> Bjorn Solberg <bjorn_ding@hekneby.org> writes:
>> When I do 5 RET to fetch the last 5 messages, I consistently only get 3,
>> missing some in-between: It's getting N-3, N-1 and N, skipping N-4 and
>> N-2.
>> 
>> If I fetch the last 6 messages, it skips N-5 in addition.
>> 
>> If I fetch the last 7 messages, it skips the same as for 6 but fetches
>> message# N-6.
>> 
>> The full *imap-log* is 800 lines long for the 5 RET, see below for the
>> last 400+ lines.  It looks like it's issuing the correct command, "UID
>> FETCH 780:784 [...]" but for some reason it doesn't display the first and
>> third of those last 5 messages.

> I didn't spot any problem in the IMAP log.

> After you have entered the group with 5 RET, what is in the " *nntpd*"
> buffer?  Note leading SPC.

> One idea is that the IMAP parser doesn't understand the responses and
> skip the article, in that case the *nntpd* buffer would only contain
> the information from three of the articles.

" *nntpd*" shows an entry for all 5 messages.  I have received new
messages since last, so now it show only 2 messages when I do 5 RET.
Which means this is somehow content-related.

However, I see a pattern in " *nntpd*" now: " *nntpd*" shows that the
messages aren't sorted by id.  For example, doing 5 RET I see entries 793
795 794 792 791 in " *nntpd*", and only 793 and 795 is shown in the
Summary buffer.  

Doing 10 RET, I see 789 790 787 786 788 793 795 791 792 794, but only 789
790 793 795 is displayed in the Summary buffer.  Which means Gnus is
looping over the entries and drops or forgets anything that looks older
than a previous encountered entry.  This would be OK if it sorted all the
entries first or something.

Despite not being too familiar with the Gnus source and being somewhat of
a Lisp beginner, I could try to track this down if I'm the only one
seeing this problem, although I suspect there are plenty of others on
this list that could point me to the right place immediately. :-) I'd
have to find out what the intention is behind the current behavior and
find a way to correct the behavior without working against the intention.

Bjorn.