Re: [PATCH] Expand explanation of the use of + in git push refspecs.

From: Marc Branchaud <marcnarc@xiplink.com>
Date: Feb 24, 2009, 16:54 UTC
Message-ID: <49A42655.40708@xiplink.com>
In-Reply-To: <7viqn5k8c4.fsf@gitster.siamese.dyndns.org>

Junio C Hamano wrote:

Show 5 quoted lines

> 
> I am of two minds with this.  The understanding of the concept of
> fast-forwardness is necessary not just to understand push but also to
> understand merge, and if glossary is missing the definition, we should add
> one there.

It's precisely because fast-forwarding and merging are so closely related (at least in my mind) that I felt the need to clarify push's documentation. My intuition about "non-fast-forward" is in the context of pulling and merging. So the current man page sounds to me like push's non-fast-forward is a merge, which it clearly isn't.

Show 12 quoted lines

> What exactly happens when the <dst> is updated is the same
> regardless of ff or non-ff in that the old 40 hexdecimal object name is
> gone and replaced with the new one, and it does not feel right to say "if
> ff, we update if non-ff you can force to overwrite."  Either way, you
> overwrite and there is no trace of the old one.  So we would want to say
> something like:
> 
>     The name of the object referenced by <src> is used to update the ref
>     <dst> on the remote side, but by default this is only allowed if the
>     update is fast-forward.  By having the optional leading plus, you can
>     tell git to update the ref <dst> even when the update is not a
>     fast-forward.

My second attempt used your wording (mostly), and also emphasized the not-merging aspect.

> Please don't wrap a full paragraph, only to change a few sentences.
> AsciiDoc wraps lines for output just fine; it took me a few extra minutes
> to make sure there is no other changes.

Whoops, sorry!  Will avoid reflowing in the future.

M.