Re: [PATCH] test-lib: avoid full path to store test results

[Jeff King <peff@peff.net>]

On Tue, Oct 30, 2012 at 05:12:57AM +0100, Felipe Contreras wrote:

> No reason to use the full path in case this is used externally.

I think it is not just "no reason to", but it is actively wrong to use a
full path, as we do not take care to "mkdir -p" the intervening path
components.

However, this never comes up in practice, because all of the test
scripts assume you are running them from the test directory (i.e.,
they will fail otherwise because they will not find ./test-lib.sh).

Is this in support of putting remote-hg tests in contrib/? I had
expected you to just put

  export TEST_DIRECTORY="$(pwd)/../../../t"
  . "$TEST_DIRECTORY/test-lib.sh"

into the test script in contrib/remote-hg/t. I guess you are doing
something like:

  cd ../../../t && ../contrib/remote-hg/t/twhatever...

but the former seems much simpler to invoke (and if the goal is to get
your test-results in the right place, setting TEST_OUTPUT_DIRECTORY is
the best way to do that).

If this is part of the remote-hg series, I'd prefer to just see it as
part of the re-roll. It's much easier to evaluate it in context.

Or are you really just doing:

  cd git/t
  $PWD/t0000-basic.sh

I guess there is nothing wrong with that, though there is no reason not
to use "./" instead of $PWD.

-Peff