Re: [Nmh-workers] Big patch: Add XOAUTH2 support for SMTP and POP

[David Levine]

Eric wrote:

> David Levine <address@hidden> writes:
>
> > test-mhlogin failed for me because each line of my actual output
> > ends with a Ctrl-M.  On Linux, so I don't know why.
>
> Did you use git am?

Yes.

> > Any idea with test-inc and test-msgcheck would fail with:
> >
> >     'inc: POP server does not support SASL'
> >     'msgchk: POP server does not support SASL'
> >
> > I did build with sasl, of course, and confirmed with mhparam.
>
> You don't need to build with SASL.  I wonder if make check
> rebuilds everything?  Maybe you need to make clean first?
> Not sure if the dependencies are right.  I'll see if I can
> reproduce this.

I think you caught it:  I ran the individual tests without
make check.  When I use that, I just get the line-ending
problem.  Also this diff, which I also saw before but didn't
mention:

    ! User-Agent: nmh/1.6+dev libcurl/7.32.0

    ! User-Agent: nmh/

> > warning: squelched 5 whitespace errors
> > warning: 10 lines add whitespace errors.
>
> They seem to fall into 3 categories:
>
> - in jsmn
>
>   I'd rather leave this pristine.

That's fine.

> - trailing space in an error message
>
>   The error message really has that trailing space on FreeBSD.
>   I forgot to highlight this issue earlier, but the 'refresh
>   finds no http server' in test-send tests an error message from
>   curl that appears on Linux but not FreeBSD.  I'm going to dig
>   into curl to see if I can figure out what's going on, but it
>   means I'm not entirely sure the test suite is correct on
>   all platforms.

I don't know that we'll ever be entirely sure.

> - trailing space in send.c that I moved to another line
>
>   removed
>
> > You chose not to use Cyrus SASL for XOAUTH2.  I wouldn't
> > hesitate to use it:  nmh already can be configured Cyrus SASL
> > and some of us do use it.  It is configured in by the Fedora
> > package.
>
> The SASL interaction here is super trivial.  It's the tiniest
> part of the patch.  I wouldn't even know where to begin to do
> what I've done with Cyrus instead.  And I'd be astonished if that
> version wasn't bigger and more complicated.

OK, I (or more likely, Ken) will look at this later.

I also wonder if there's a way to have TLS connection such that
the proxy wouldn't be needed.  We've already got it for outgoing
with send/post -initialtls.

Do you have commit access to the nmh git repo?  This could/
should go in on a branch.

David