Re: [PATCH v2 2/2] Doc/check-ref-format: clarify information about @{-N} syntax

From: Kaartic Sivaraam <kaartic.sivaraam@gmail.com>
Date: Nov 28, 2017, 14:43 UTC
Message-ID: <1511880237.10193.5.camel@gmail.com>
In-Reply-To: <xmqqd14386sw.fsf@gitster.mtv.corp.google.com>

On Tue, 2017-11-28 at 11:40 +0900, Junio C Hamano wrote:

Show 10 quoted lines

> Kaartic Sivaraam <kaartic.sivaraam@gmail.com> writes:
> 
> > When the N-th previous thing checked out sytax is used with
> > '--branch' option of check-ref-format the results might not
> > always be a valid branch name
> 
> I wonder if you want to rephrase this, because 40-hex object name is
> syntactically a valid branch name.  It's (1) cumbersome to type and
> (2) may not be what the user expects.
>

You're right. Actually a previous draft of that log message did say something like,

	Though a commit-hash might be a valid branch name, it isn't
	something that's expected by the users of "check-ref-format".

I removed as I thought it would be unnecessary. It seems I took the wrong decision. Will fix it. :-)

Show 5 quoted lines

> I have a mild suspicion that "git checkout -B @{-1}" would want to
> error out instead of creating a valid new branch whose name is
> 40-hex that happen to be the name of the commit object you were
> detached at previously.
>

I thought this the other way round. Rather than letting the callers error out when @{-N} didn't expand to a branch name, I thought we should not be expanding @{-N} syntax for "check-ref-format --branch" at all to make a "stronger guarantee" that the result is "always" a valid branch name. Then I thought it might be too restrictive and didn't mention it. So, I dunno.

> I am not sure if "check-ref-format --branch" should the same; it is
> more about the syntax and the 40-hex _is_ valid there, so...

I'm not sure what you were trying to say here, sorry.

-- 
Kaartic