temporarily unsetting sendemail.smtpuser

temporarily unsetting sendemail.smtpuser

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 1717 bytes --]

I have a setup where I normally send patches through my ISP, which
requires authentication, so I have sendemail.smtpuser=me and
sendemail.smtpserverport=587 set in my ~/.gitconfig (among others).  But
for one project I work on, I am required to send patches through an
alternate SMTP server that does not require authentication.

At first, I thought it would be a simple matter to use git config to
create an alias, where the explicit command line requests would override
the global ~/.gitconfig settings:

[alias]
	submit = send-email --smtp-user= --smtp-server=alternate \
		--smtp-server-port=25

But that fails with:

$ git submit -1
fatal: unrecognized argument: --smtp-user=
format-patch -o /tmp/JgTFO8Rq3v -1 --smtp-user=: command returned error: 128

In other words, git-send-email didn't recognize the empty --smtp-user=
command as an override request, and instead passed it on to
git-format-patch which croaks.

So I ended up creating this hairy alias instead:

[alias]
	submit = "!sh -c 'git config --global --unset sendemail.smtpuser; git
send-email --smtp-server=alternate --smtp-server-port=25 \"$@\"; st=$?;
git config --global sendemail.smtpuser me; (exit $st)' sh"

I suppose I could have used a shell function for fewer forks:

[alias]
	submit = "!doit() { git config --global --unset sendemail.smtpuser; git
send-email --smtp-server=alternate --smtp-server-port=25 \"$@\"; st=$?;
git config --global sendemail.smtpuser me; return $st; }; doit"

But both of those are rather unappealing.  Is there a better approach to
accomplishing what I wanted?

-- 
Eric Blake   eblake@redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 619 bytes --]

Re: temporarily unsetting sendemail.smtpuser

[Ævar Arnfjörð Bjarmason]

On Wed, May 19, 2010 at 22:54, Eric Blake <eblake@redhat.com> wrote:
> But both of those are rather unappealing.  Is there a better approach to
> accomplishing what I wanted?

--no-smtp-user ?