Re: [PATCH/RFC 2/2] git rebase -i: Warn removed or dupplicated commits

From: Junio C Hamano <gitster@pobox.com>
Date: May 28, 2015, 16:53 UTC
Message-ID: <xmqqfv6giqyu.fsf@gitster.dls.corp.google.com>
In-Reply-To: <1388345544.70438.1432799047393.JavaMail.zimbra@ensimag.grenoble-inp.fr>

Remi Galan Alfonso <remi.galan-alfonso@ensimag.grenoble-inp.fr> writes:

Show 14 quoted lines

> Junio C Hamano <gitster@pobox.com> writes:
>> I think there is a difference between (silently) accepting just to
>> be lenient and documenting and advocating mixed case uses.
>> 
>> Personally, I'd rather not to see gratuitous flexibility to allow
>> the same thing spelled in 47 different ways for no good reason.
>
> It was more of a mistake on our part rather than actually wanting to
> document mixed case uses.
>
> In the v2 of the patch (not sent to the mailing list yet since we want
> to take into consideration the conclusion of this discussion before)
> it is entirely in lower case in both the documentation and the code
> while we silently allow upper and mixed case.

Understood; I am not sold on the whole "warning" business, though.

I think I saw you did 'tr [:upper:]' or something like that in the patch; we tend to avoid [:class:] and [=equiv=] when not needed, unless we know that the matching engine used supports them (i.e. it is OK to use them in Perl scripts and it is OK to feed them to the wildmatch-based matcher in Git itself, but not in general shell scripts). As the values can all be represented in US-ASCII, it should be sufficient to do "tr 'A-Z' 'a-z'", I would think.