Continuing struggle with expiry

Continuing struggle with expiry

[Norman Walsh]

Sigh. I can't tell if I'm making progress or not. By a process of
mostly shotgun debugging, I now have a bunch of articles in my spam
folder that are marked as expirable.

The group parameters are:

((timestamp 15690 40431)
 (auto-expire . t)
 (total-expire . t)
 (expiry-wait . 7))

When I enter the summary buffer with C-u space, I can see that there
are a lot of 'E' articles and a lot of them are way older than 7 days.

So it seems to me that some process should be actually deleting those
old, expired articles. But that doesn't happen when I leave the
summary buffer.

What function actually cleans up articles marked expired that are older
than expiry-wait days? What hook/function/magic runs that function?

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | Nothing will ever be attempted, if all
http://nwalsh.com/            | possible objections must be first
                              | overcome.--Dr. Johnson

Re: Continuing struggle with expiry

[Paul Jarc]

Norman Walsh <ndw@nwalsh.com> wrote:
> What function actually cleans up articles marked expired that are older
> than expiry-wait days?

nn*-request-expire-articles.

> What hook/function/magic runs that function?

debug-on-entry your n-r-e-a function and look at the call stack.  If
the debugger isn't entered, then there's a problem upstream; grep the
source.


paul

Re: Continuing struggle with expiry

[Kai Großjohann]

Norman Walsh <ndw@nwalsh.com> writes:

> What function actually cleans up articles marked expired that are older
> than expiry-wait days? What hook/function/magic runs that function?

gnus-summary-expire-articles is normally run from
gnus-summary-exit-hook, I think.

[looking]

Oh, it's gnus-summary-prepare-exit-hook.

kai
-- 
A large number of young women don't trust men with beards.  (BFBS Radio)

Re: Continuing struggle with expiry

[Norman Walsh]

/ prj@po.cwru.edu (Paul Jarc) was heard to say:
| Norman Walsh <ndw@nwalsh.com> wrote:
|> What function actually cleans up articles marked expired that are older
|> than expiry-wait days?
|
| nn*-request-expire-articles.

Uncle! I like to think I enjoy elisp hackery as much as the next guy,
but the more I investigate this problem, the less sense it makes.

Expiry isn't working for me because nnml-request-expire-articles calls
nnmail-expired-article-p with mod-time set to '(15679 803). And when
nnmail-expired-article-p asks for the (time-since '(15679 803)), time-since
goes bang:

Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p nil)
  time-since((15679 803))
  eval((time-since (quote (15679 803))))

That is, the byte-compiled, auto-loaded version of time-since goes
bang. If I copy the definition of time-since into *scratch* and
compile it myself, which I did initially in order to debug it further,
it runs just fine, thank you very much.

BTW, "goes bang" isn't strictly true. nnmail-expired-article-p calls
(ignore-errors (time-less-p ...)) so the error in time-since is
ignored, and nil is returned, meaning that no article is expired.

Help!?

For the record, I'm running GNU Emacs 21.1.1 and ognus from CVS this morning.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | We think in generalities, but we live
http://nwalsh.com/            | in detail--Alfred North Whitehead

Re: Continuing struggle with expiry

[Simon Josefsson]

Norman Walsh <ndw@nwalsh.com> writes:

> / prj@po.cwru.edu (Paul Jarc) was heard to say:
> | Norman Walsh <ndw@nwalsh.com> wrote:
> |> What function actually cleans up articles marked expired that are older
> |> than expiry-wait days?
> |
> | nn*-request-expire-articles.
>
> Uncle! I like to think I enjoy elisp hackery as much as the next guy,
> but the more I investigate this problem, the less sense it makes.
>
> Expiry isn't working for me because nnml-request-expire-articles calls
> nnmail-expired-article-p with mod-time set to '(15679 803). And when
> nnmail-expired-article-p asks for the (time-since '(15679 803)), time-since
> goes bang:
>
> Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p nil)
>   time-since((15679 803))
>   eval((time-since (quote (15679 803))))
>
> That is, the byte-compiled, auto-loaded version of time-since goes
> bang. If I copy the definition of time-since into *scratch* and
> compile it myself, which I did initially in order to debug it further,
> it runs just fine, thank you very much.

Are you using the correct time-since?  I think a problem like this was
solved in time-date.el a few months ago, so if you have an older byte
compiled version still laying around somewhere, perhaps your emacs
uses the wrong one.  Try C-h f time-since in a fresh Gnus.

Re: Continuing struggle with expiry

[Norman Walsh]

/ Simon Josefsson <jas@extundo.com> was heard to say:
| Are you using the correct time-since?  I think a problem like this was
| solved in time-date.el a few months ago, so if you have an older byte
| compiled version still laying around somewhere, perhaps your emacs
| uses the wrong one.  Try C-h f time-since in a fresh Gnus.

Thank you. I discovered a broken implementation of time-seconds in
make-regexp.el. I use make-regexp in my .gnus, so naturally it got
loaded first.

Whew!

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | Young men think old men are fools; but
http://nwalsh.com/            | old men <em>know</em> young men are
                              | fools.--George Chapman

Re: Continuing struggle with expiry

[Kai Großjohann]

Norman Walsh <ndw@nwalsh.com> writes:

> Thank you. I discovered a broken implementation of time-seconds in
> make-regexp.el. I use make-regexp in my .gnus, so naturally it got
> loaded first.

This only goes to show that packages should _never_ define a function
in the global namespace.

kai
-- 
A large number of young women don't trust men with beards.  (BFBS Radio)