nnfolder groups losing expirable marks

nnfolder groups losing expirable marks

[Dan Christensen]

Using the Jan 4 version of git gnus (a1d4c9b50...) I'm finding that my
expirable marks are very frequently being lost in nnfolder groups.  It
seems to happen whenever expiry is run:  the articles that aren't old
enough to be expired lose their expirable marks.  But I haven't done
enough experimenting to be sure this is the pattern.

I checked the log of later commits and didn't see anything relevant,
but I am happy to upgrade to the latest version if that would help.
Or to try to do a bisection.

Dan

Re: nnfolder groups losing expirable marks

[Dan Christensen]

Dan Christensen <jdc@uwo.ca> writes:

> Using the Jan 4 version of git gnus (a1d4c9b50...) I'm finding that my
> expirable marks are very frequently being lost in nnfolder groups.  It
> seems to happen whenever expiry is run:  the articles that aren't old
> enough to be expired lose their expirable marks.  But I haven't done
> enough experimenting to be sure this is the pattern.

I have now experimented some more, and this seems to be exactly the
pattern.  It is 100% reproducible in nnfolder groups, by marking recent
articles as expirable and then running gnus-group-expire-articles on
the group.

> I checked the log of later commits and didn't see anything relevant,
> but I am happy to upgrade to the latest version if that would help.

I checked today's latest version and the problem is still there.
Moreover, I figured out that the problem starts exactly with this
commit:

commit 06eeefe0c892fd77f7a79493fa4f1db0f8058761
Author: Lars Magne Ingebrigtsen <larsi@quimbies.gnus.org>
Date:   Thu Dec 16 18:58:19 2010 +0100

    Clear up nnml expiry target performance regressions.
    
    Fix previous fix by having nnfolder save all its buffers
    explicitly.

I'm hesitant to change this, though, since it seems this was done for a
reason.  And I'm not sure what "previous fix" is being referred
to---probably it means the nnml fix.

In case it is relevant, I have nnfolder-marks-is-evil set to t and the
groups where I've noticed this are auto-expire groups.

Dan

Re: nnfolder groups losing expirable marks

[Lars Ingebrigtsen]

Dan Christensen <jdc@uwo.ca> writes:

> I have now experimented some more, and this seems to be exactly the
> pattern.  It is 100% reproducible in nnfolder groups, by marking recent
> articles as expirable and then running gnus-group-expire-articles on
> the group.

Hm.  I've looked at the changeset you narrowed it down to, and I can't
really understand how that is leading to the expiry marks being
removed.  Does this only happen in nnfolder groups?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnfolder groups losing expirable marks

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Hm.  I've looked at the changeset you narrowed it down to, and I can't
> really understand how that is leading to the expiry marks being
> removed.  Does this only happen in nnfolder groups?

I found the bug now.  The function is supposed to return the list of
unexpired articles, which it no longer did.  Trying to push out the fix
now if I can get my git-fu working here...

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnfolder groups losing expirable marks

[Dan Christensen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> I found the bug now.  The function is supposed to return the list of
> unexpired articles, which it no longer did.  Trying to push out the fix
> now if I can get my git-fu working here...

Yay, it works!  Thanks.

Dan