From: Junio C Hamano <gitster@pobox.com>
Date: Sat, 07 May 2011 18:56:46 GMT
Subject: Re: [PATCH] t3703: add test cases for pathspec magic
Message-ID: <7vwri2wd8x.fsf@alter.siamese.dyndns.org>
In-Reply-To: <1304764507-27547-1-git-send-email-pclouds@gmail.com>

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> That was the intention, but it raises a question: what do we do if a
> file happens to have the same name with the given magic pathspec, as
> in the last two tests?

I would have expected that in such a case the user would pass a "\:" or
even ":::" to match the file ":", and "\:(rubbish)" to match the file
":(rubbish)".  The whole ":" is special thing comes from the observation
that a path that begins with a colon is rare, so it is Ok to require the
user to do some more work (typing an extra backslash) when he really wants
to match with such a thing.

A script that takes a pathname that is meant to be a literal from the user
in its variable $x would pass ":(noglob)$x" when it wants to be strict.  A
script that lets the user say whatever and wants to pass would just pass
"$x" along the callchain.

I do not expect this to be an issue in practice, though.  Have you seen a
script that tries to quote all the possible globbing characters in "$x"
before calling into git with the current codebase without this magic?