Re: [PATCH] t6024-recursive-merge.sh: hide spurious output when not running verbosely

[Jeff King <peff@peff.net>]

On Fri, Feb 29, 2008 at 03:50:03PM -0800, Junio C Hamano wrote:

> Actually, I think this might be a bit more sensible approach.
> 
> -- >8 --
> tests: allow optional clean-up phrase to expect_success/failure
> 
> When one test modifies the state of the test repository that the later
> tests may depend on, you may want to add a clean-up action that is run
> regardless of the outcome of the main part of the test.
> 
> This can now be specified as the third parameter to test_expect_success
> and test_expect_failure functions.

I think your heart is in the right place with this patch, but I doubt
that it is going to be all that productive in practice. Most tests
consist of a long list of commands, and cleaning up properly after every
possible failure case is going to be a lot of work. And worse, since the
tests generally _don't_ fail, you have no way to test that your cleanup
is reasonable.

So I think we will end up in the case where a few failed tests will
properly clean themselves up and let further tests proceed, but most
failures will leave a big question. In other words, what problem have we
solved?  If tests N and N+k both fail, would you, even with this patch,
suspect N+k of actually failing, or would you first go and debug test N?
Is that any different than what you do now?

-Peff