From: Junio C Hamano <gitster@pobox.com>
Date: Mon, 20 Jul 2009 21:34:09 GMT
Subject: Re: [PATCH RFC 2/8] push: add push.default = mirror
Message-ID: <7vhbx7ujhq.fsf@alter.siamese.dyndns.org>
In-Reply-To: <4A64DE38.5090805@gmail.com>

Paolo Bonzini <paolo.bonzini@gmail.com> writes:

>> I think this patch alone (regardless of the rest which I haven't read)
>> probably makes sense.
>>
>> Except that I think the part below contradicts with the --mirror push
>> semantics (see remote.c::match_refs()).
>
>        if (!send_mirror && prefixcmp(src->name, "refs/heads/"))
>                 continue;
>
> The way I read this, send_mirror will send everything, even if it's
> outside refs/heads/*.  So refs/*.

Ah, ok, thanks.

> If we agree on the semantics, would you harvest it separately or
> should I resubmit?

I do not expect the rest of the series to be near 'next' before 1.6.4, and
it is likely that I'll forget about this patch by the time 1.6.4 finally
ships.

But I realize that without the rest of the series, being able to say
push.default = mirror does not add much value.  A replacement commit log
message I wrote for this patch stand-alone is:

    Instead "git push --mirror $over_there", you can say "git push" after
    setting up push.default to "mirror".

But the thing is, you can already do that with "remote add --mirror".