On 10/03/2017 20:42, Jeff King wrote:
That's something I guess, but I'm not enthused by the idea of just dumping a bunch of binary test cases that nobody, not even the author, understands.
[...]
My real concern is that this is the tip of the ice berg. So we increased coverage in one program by a few percent. But wouldn't this procedure be applicable to lots of _other_ parts of Git, too?
I think that index-pack is in a special position given its role as the verifier for packs received over the network, which you also wrote here: https://www.spinics.net/lists/git/msg265118.html I also think increased coverage for other parts of git which are not considered security-sensitive is less valuable without testing for an actual expected result. Vegard