On Wed, May 16, 2018 at 09:56:33AM -0700, Stefan Beller wrote: > Hi brian, > > On Tue, May 15, 2018 at 6:56 PM, brian m. carlson > wrote: > For the 2 occurrences above I think the SHA1 requirement is not > needed as the check if a blob exists (and the id is given as $1) > is independent of the hash function, it is just important that > the same hash function is used in the git-cat-file as well as... > > ... here, where we create the blob to test without > writing it into the object database. In a way we test that > the absence of -w works correctly. > > Oh, the $hello_sha1 is hard coded, which is why we > think this test is SHA1 dependent. > > But that would fit in line with the test_blob[_does_not]_exist > being independent of the hashes? These functions are technically independent of the hash, but the way we call them is not. Since we only look up certain fixed values in those functions, they're going to fail if we use a different hash. There really isn't a great way to annotate the tests independent of the functions without duplicating a lot of the logic that occurs in the test library, and I didn't really want to do that. -- brian m. carlson: Houston, Texas, US OpenPGP: https://keybase.io/bk2204