Re: [RFC/PATCH] pkt-line: allow writing of LARGE_PACKET_MAX buffers

On Tue, Dec 09, 2014 at 08:58:26PM +0100, Johannes Sixt wrote:

Am 09.12.2014 um 18:49 schrieb Jeff King:

quoted

+test_expect_success 'create repo with absurdly long refname' '
+	ref240=$_z40/$_z40/$_z40/$_z40/$_z40/$_z40
+	ref1440=$ref240/$ref240/$ref240/$ref240/$ref240/$ref240 &&
+	git init long &&
+	(
+		cd long &&
+		test_commit long &&
+		test_commit master &&
+		git update-ref refs/heads/$ref1440 long

Having this ref on the filesystem is going to fail on Windows, I presume
(our limit is 260). Can we stuff it away as a packed ref right from the
beginning? (And turn off reflogs, BTW.)

Yeah, after sending that, I wondered if it would cause problems.

I don't think it would be too hard to just cat it right into the
packed-refs file. The other option would be to try creating it, and set
a prereq to skip other tests if it fails.

-Peff