Re: failure using starttls with Oort gnus/imap

Re: failure using starttls with Oort gnus/imap

[Lars Magne Ingebrigtsen]

William F Hammond <hammond@csc.albany.edu> writes:

>   rename-buffer(#<killed buffer>)
>   imap-open("imap.albany.edu" nil nil nil #<killed buffer>)

It should probably be passed the buffer name, and not the buffer
itself.  Fix in CVS.

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

Re: failure using starttls with Oort gnus/imap

[Lars Magne Ingebrigtsen]

William F Hammond <hammond@csc.albany.edu> writes:

> After moving to GNU Emacs 21.3. and using the overnight tarball
> dated 20030320, I get a somewhat different debug message, which matches
> byte-for-byte the debug message from ognus-0.16:

The problem seems to be more fundamental.  `imap-open' may kill the
buffer it's been fed, for some reason or other, which makes
`mail-source-fetch-imap' very unhappy, since it created the buffer.

Perhaps Simon knows what the right fix for this should be?  (I think
it's probably a mistake for `imap-open' to kill the buffer...)

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

Re: failure using starttls with Oort gnus/imap

[Simon Josefsson]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> William F Hammond <hammond@csc.albany.edu> writes:
>
>> After moving to GNU Emacs 21.3. and using the overnight tarball
>> dated 20030320, I get a somewhat different debug message, which matches
>> byte-for-byte the debug message from ognus-0.16:
>
> The problem seems to be more fundamental.  `imap-open' may kill the
> buffer it's been fed, for some reason or other, which makes
> `mail-source-fetch-imap' very unhappy, since it created the buffer.
>
> Perhaps Simon knows what the right fix for this should be?  (I think
> it's probably a mistake for `imap-open' to kill the buffer...)

imap.el connects to the server initially, and then looks at what
authentication/stream mechanisms are available ("starttls", "imtest",
etc), and then tries each one of them in order until someone succeeds.
These subsequent connections are made in a temporary buffer.  If one
of them succeeds, the first buffer is closed and the new one renamed
to the old one.

This works fine in Gnus which usually uses name of buffers, but I
guess mail-sources use the elisp buffer type instead.

The reason the initial buffer isn't teared down and the other streams
are invoked in the same buffer, is that imap.el can (and frequently
does) decide to use that connection if no stronger
stream/authentication succeeded.  So if it is teared down, it would
have to reopen it.

I'll see if I can think of a way to fix it...

Re: failure using starttls with Oort gnus/imap

[William F Hammond]

Simon Josefsson <jas@extundo.com> writes:

> Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
. . .
> > The problem seems to be more fundamental.  `imap-open' may kill the
> > buffer it's been fed, for some reason or other, which makes
> > `mail-source-fetch-imap' very unhappy, since it created the buffer.
> >
> > Perhaps Simon knows what the right fix for this should be?  (I think
> > it's probably a mistake for `imap-open' to kill the buffer...)
> 
> imap.el connects to the server initially, and then looks at what
> authentication/stream mechanisms are available ("starttls", "imtest",
> etc), and then tries each one of them in order until someone succeeds.
> These subsequent connections are made in a temporary buffer.  If one
> of them succeeds, the first buffer is closed and the new one renamed
> to the old one.
> 
> This works fine in Gnus which usually uses name of buffers, but I
> guess mail-sources use the elisp buffer type instead.
> 
> The reason the initial buffer isn't teared down and the other streams
> are invoked in the same buffer, is that imap.el can (and frequently
> does) decide to use that connection if no stronger
> stream/authentication succeeded.  So if it is teared down, it would
> have to reopen it.

Many thanks to both of you.

I just want to point out three things:

  0.  Very, very small messages seem to get through.

  1.  The same set-up talking to the same imap server (cyrus) for the
      same account from a GNU/Linux host works fine.

  2.  A mail-sources set-up that differs only in the mail-sources spec
      having

                :stream    ssl

      works fine on the same SunOS 5.8 host.

So I'm inclined to guess that at the base some library call by
starttls is not doing what is expected.  Unfortunately I'm not in a
position to give starttls (v. 0.9) a free-standing check out.

Thanks again.

                                    -- Bill