Re: [PATCH v3 2/4] cat-file: introduce the --filters option

From: Junio C Hamano <gitster@pobox.com>
To: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Cc: git@vger.kernel.org, "Torsten Bögershausen" <tboegi@web.de>,
	"Jeff King" <peff@peff.net>
Subject: Re: [PATCH v3 2/4] cat-file: introduce the --filters option
Date: Fri, 09 Sep 2016 09:08:10 -0700	[thread overview]
Message-ID: <xmqqbmzxdpjp.fsf@gitster.mtv.corp.google.com> (raw)
In-Reply-To: <alpine.DEB.2.20.1609091800020.129229@virtualbox> (Johannes Schindelin's message of "Fri, 9 Sep 2016 18:01:21 +0200 (CEST)")

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

>> > +	if (type != OBJ_BLOB) {
>> > +		free(*buf);
>> > +		return error(_("blob expected for %s '%s'"),
>> > +			sha1_to_hex(sha1), path);
>> > +	}
>> > +	if (S_ISREG(mode)) {
>> > +		struct strbuf strbuf = STRBUF_INIT;
>> > +		if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
>> > +			free(*buf);
>> > +			*size = strbuf.len;
>> > +			*buf = strbuf_detach(&strbuf, NULL);
>> > +		}
>> > +	}
>> 
>> This needs to error out when mode is not ISREG just like it errors
>> out when type is not BLOB.
>
> Are you sure that this is desirable in batch mode?

I do not quite see a reason why we should not diagnose a bad input
that does not produce a filtered result.  In batch mode or not, it
diagnoses when the user feeds it a non-blob, and I think it should
do so for non-regular, too. Both are "you asked me to filter, but
you shouldn't have".