Re: What's cooking in git.git (May 2013, #05; Mon, 20)

From: Junio C Hamano <gitster@pobox.com>
Date: May 23, 2013, 17:37 UTC
Message-ID: <7vli75fvh4.fsf@alter.siamese.dyndns.org>
In-Reply-To: <519E366A.5040504@drmicha.warpmail.net>

Michael J Gruber <git@drmicha.warpmail.net> writes:

> Well, if we decide "showing blobs with textconv is fundamentally
> different from showing diffs with textconv" then "--textconv" should not
> apply any textconv filters on blobs unless the user has specified them
> using a separate attribute (different from "diff").

I had an impression that the ship has already sailed wrt to "diff" being pretty much interchangeable as "text" (or "non binary") in the attribute system long time ago.

Show 5 quoted lines

> Therefore, I hesitate introducing the behavior of the current series.
> For me, it would introduce something of a "mixed beast".
>
> I wouldn't hesitate introducing "textconv on by default for blobs the
> same as for diffs",

I would. But I wouldn't for "the user asks for --textconv, the user expects the blob shown with mangling", which sounds like a good thing to do.

And there is anything wrong to refine later where exactly that textconv filter is defined. At this moment, "diff.textconv" is the only place the user could even contemplate setting one, and there is no risk for confusion. "blob.textconv" can be introduced much later when some users actually want to have a pair of different filters, at that point "diff.textconv" will become a backward compatiblity fallback for that filter.