Re: What's cooking in git.git (Mar 2019, #04; Wed, 20)

[Junio C Hamano <gitster@pobox.com>]

Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:

>     	graph_version = *(unsigned char*)(data + 4);
>     <<<<<<< HEAD
>     	if (!graph_version || graph_version > 2) {
>     		error(_("unsupported graph version %X"),
>     		      graph_version);
>     =======
>     	if (graph_version != GRAPH_VERSION) {
>     		error(_("commit-graph version %X does not match version %X"),
>     		      graph_version, GRAPH_VERSION);
>     		return NULL;
>     	}
>
>     	hash_version = *(unsigned char*)(data + 5);
>     	if (hash_version != oid_version()) {
>     		error(_("commit-graph hash version %X does not match version %X"),
>     		      hash_version, oid_version());
>     >>>>>>> commit-graph-fix-segfault-and-exit-3
>     		return NULL;
>     	}
>
> Needs to be resolved as:
>
> 	graph_version = *(unsigned char*)(data + 4);
> 	if (!graph_version || graph_version > 2) {
> 		error(_("commit-graph the graph version %X is unsupported"),
> 		      graph_version);
> 		return NULL;
> 	}
>
> I.e. there's a test that greps out "graph version".

Yikes.

Given the common ancestor version's phrasing, and also the updated
phrasing of the other message since we started supporting v2 of the
commit-graph file, I resolved this message to 

	 "unsupported commit-graph version %X"

instead.  Of course, I wasn't expecting the test to be depending on
the exact error message's phrasing X-<.

Thanks.