Re: Moving from nnml to nnimap loses marks?

[Simon Josefsson <jas@pdc.kth.se>]

Kai.Grossjohann@CS.Uni-Dortmund.DE writes:

> Search for the string "kai" to find the places where I added code.
> From looking at it, this seems to be meaningful code

Yes, this was exactly the patch I had in mind.

> I created a group nnml:test with three messages in it.  The first
> message is marked "O", the second and third are marked "!".  I
> copied these messages to a (previously empty) group
> nnimap:user.grossjoh.test, then had a look at it, and the second
> message was unmarked (the others had kept their marks).  I then
> deleted all messages from the nnimap group and did the same thing
> again.  This time, the third message was unmarked, and the other two
> had again kept their marks.

I wasn't able to repeat this, for me your patch work fine.

Your patch makes "ticked" articles not have the "read" mark though,
Gnus want a read mark on ticked articles. This only result in a
cosmetic miscount on the number of unread articles in a folder and
gnus/nnimap also detect and correct this when you enter the mailbox
next time (as seen in your debug logs).

I placed a slightly modified version of your patch on the patch page.

> The following from the *Messages* buffer shows that
> gnus-request-set-mark was called with the right arguments:

Yup. Seems all right.
 
> I wonder where are the nnimap-request-set-mark calls coming from where
> the action is to delete the mark?

It's a known but harmless bug, but you can see that it only happens
after you've deleted the articles in question. So it only want to
propagate that Gnus has removed the article (hence all flags on the
article) to the server.

> Anyway, even though this is lots of information, I hope it will help
> someone understand what's going on.

I went through your trace and I couldn't find anything wrong with
it. I did the same as you (created a empty nnml group, copied three
articles from it to a empty nnimap group and inspected the flags) and
I was able to repeat it successfully several times.

Hm. I did only try my slightly modified patch, but I don't think
thoose modifications should affect anything. Anyway your traces
doesn't indicate that anything is wrong.

Could you `G E' "nnimap:user.grossjoh.test" after you've copied the
articles to it, but before you enter the group, and compare that with
how it looks after you've entered the group? That way we know what
Gnus believes about the group.

Could it perhaps be some sorting/threading that changes order of the
articles so it looks like they have changed flags?

-s