From: Junio C Hamano Date: Mon, 25 Apr 2011 19:31:24 GMT Subject: Re: [PATCH] t/README: unify documentation of test function args Message-ID: <7vzknei0vn.fsf@alter.siamese.dyndns.org> In-Reply-To: <4DB5BDBE.308@debugon.org> Mathias Lafeldt writes: > Ah, okay. Is this true for test_expect_code too? It also has a git command > in the example, but it says , not : > > - test_expect_code I think that one is Ok as there is no git specific hacks in the semantics. It is a straight "Run this command, and make sure you get this exit code". The "must-fail" is really a special case in that we just do not expect any failure, but expect a controlled failure. Currently we only catch segv, but we should anticipate that we will reject other types of uncontrolled failures in the future, and keeping the function "git-only" makes it easier to manage.