From: Junio C Hamano <gitster@pobox.com>
Date: Sun, 26 Apr 2015 04:32:32 GMT
Subject: Re: [PATCH v2] t0027: Add repoMIX and LF_nul
Message-ID: <xmqqbnibzezj.fsf@gitster.dls.corp.google.com>
In-Reply-To: <553C0A9B.7060407@web.de>

Torsten Bögershausen <tboegi@web.de> writes:

>> Hmph, would it still make sense to make sure that CRLF will stay CRLF
>> so that future changes to break this expectation can be caught?  Not
>> that such a breakage is likely...
>
> Thanks for amending.
>
> We have the file CRLF (and CRLFmixLF), where we check that CRLF stays CRLF and is not
> converted into CRLFLF.
>
> The LF_nul is to test the "auto text detection":
> It should not be converted into CRLF_nul in "auto mode",
> but should be converted when declared as "text" in .gitattributes.

The two paragraphs above may explain why the test would pass with
the current code, but is that sufficient assurance that other people
in future changes would not break the "It should not be" with their
bugs that you do not foresee at this moment?  Whenever you say "it
should not be", I hear "if the code somehow does that, somebody
broke our code", which in turn suggests me that catching that
condition with tests is a good thing.

Or am I hearing you incorrectly?