Re: gnus-range-difference((nil . 95923) nil) → (wrong-type-argument number-or-marker-p nil)

[Kevin Brubeck Unhammer <unhammer@fsfe.org>]

[-- Attachment #1.1: Type: text/plain, Size: 1758 bytes --]

"B.V. Raghav" <bvraghav@iitk.ac.in> čálii:

> Kevin Brubeck Unhammer <unhammer@fsfe.org> writes:
>
>>
>> "commits" is a dovecot IMAP folder, synced by offlineimap to
>> fastmail. My workaround is to delete
>>
>> ("commits" (3 . 8) ((1 . 95923)) ((unexist (1 . 75894) (75896... [SNIPPED]
>>
>> From .newsrc.eld, but it keeps coming back. I've also tried deleting
>> "commits" from .newsrc, to no avail. It seems like a bug in gnus (though
>
> Thanks Kevin for the direction.
>
> In my case, it is the another remote imap server `nnimap:r-bvr' and one
> of the folder pops out with an error in the debugger. I do not know
> why, or cant predict which one?
>
> I took the deleted part from .newsrc.eld in two subsequent runs, and
> ediff-ed it. Following were the observations (n, m, a1, a2, b all are
> numbers):
>
> 1. there is a difference in the `modseq' cdr. For example,
>    (modseq . "n") changes to (modseq . "m")
>    or
>    (modseq . "n") changes to (modseq) - or vice versa

Yes, for me the working .eld has (modseq) while the current
nonworking one has (modseq . "99209").

> 2. there is a difference in active cdr. For example,
>    (active a1 . b) changes to (active a2 . b)

For me, the difference is in the b here (and only the number is
different, not the structure), so I believe the active change may be
unrelated.


But I also see a difference in the first two elements after after the
string "commits":

working .newsrc.eld has
 ("commits" 3 ((1 . 96052)) …)

failing .newsrc.eld has
 ("commits" (3 . 1) ((1 . 96052)) …)

Note also that the backtrace said it failed at
 gnus-range-difference((nil . 95923) nil)
 nnimap-update-info(("commits" (3 . 8) ((1 . 95923)) …)


-Kevin

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 472 bytes --]