Re: Performance problem of imap move

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> Lars Ingebrigtsen <larsi@gnus.org> writes:
>
>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>> A local dovecot server, synced with gmail via isync.
>>
>> Ah.  For some reason, it seems that people running local IMAP servers
>> triggers a lot of bugs in Gnus...
>>
>>> I've always meant to go and follow the code, but it's never been
>>> annoying enough for me to actually do it. But if you're messing nnimap
>>> at the moment anyway, maybe now's a good time.
>>
>> Please do.  :-)
>
> Doing this bit by bit, first of all here are the imap commands resulting
> from moving article 39339 from INBOX to [Gmail].Trash, where it became
> article 175. Upon leaving the INBOX group, [Gmail].Trash was marked as
> containing 36 unread articles, when in fact it had none -- refreshing
> Gnus reset the unread count to the proper value. Stepping through
> `gnus-summary-move-article' and `nnimap-request-move-article' didn't
> show anything immediately out of the ordinary, but I don't really
> understand imap all that well.
>
> In a bit, I'll try it again and step through `nnimap-request-set-mark'.
> That's what's being used to update marks on [Gmail].Trash, and seems
> like the next most likely place to find bugs. Like I said, though, I
> don't know imap all that well, so someone tell me if I should be looking
> elsewhere.

A little progress: `nnimap-request-set-mark' also looked normal, so I
followed the `gnus-group-get-new-news-this-group' call at the end of
`gnus-summary-move-article' instead.

In the `let' in the middle of that (gnus-group.el:4081 in current Gnus),
we call both `gnus-get-info' and `gnus-active' on the "group" variable.
The return value of `gnus-get-info' indicates a "seen" range of (1 .
177), which is correct. The return value of `gnus-active' is (1 . 211),
which isn't correct, and leads to the bogus unread count.

Off to look at what's happening in `gnus-active'...

E