Re: [PATCH v3] remote: add get-url subcommand

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 5, 2015, 21:39 UTC
Message-ID: <xmqq1tfh2z5g.fsf@gitster.dls.corp.google.com>
In-Reply-To: <20150805213358.GA2592@megas.kitware.com>

Ben Boeckel <mathstuf@gmail.com> writes:

Show 28 quoted lines

> On Wed, Aug 05, 2015 at 13:34:18 -0700, Junio C Hamano wrote:
>> Changes to these two files look reasonable.
>> 
>> Don't you want to protect this feature from future breakage by
>> others by adding a couple of tests, though, to t/t5505?
>
> Thanks, I've done so locally. It actually brings up this case:
>
>     $ git remote add someremote foo
>     $ git remote get-url --push someremote
>     fatal: no URLs configured for remote 'someremote'
>
> Is it better to use:
>
>     remote = remote_get(remotename);
>     remote->pushurl;
>
>     if (remote->pushurl_nr)
>         remote->pushurl;
>     else
>         remote->url;
>
> or:
>
>     remote = pushremote_get(remotename);
>     remote->pushurl;
>
> ? What is the actual difference between the two?

You tell me ;-)

The default remote based on the current branch is computed differently based on the direction of the transfer, I think.

        struct remote *remote_get(const char *name)
        {
                return remote_get_1(name, remote_for_branch);
        }

        struct remote *pushremote_get(const char *name)
        {
                return remote_get_1(name, pushremote_for_branch);
        }

When you are not giving name explicitly, the second parameter to _1 function is used to determine the name.