From: Jeff King <peff@peff.net>
Date: Sun, 24 May 2009 16:43:30 GMT
Subject: Re: [PATCH] t5500-fetch-pack: ignore output from git init
Message-ID: <20090524164329.GC22792@coredump.intra.peff.net>
In-Reply-To: <1243107282-9259-1-git-send-email-bebarino@gmail.com>

On Sat, May 23, 2009 at 12:34:42PM -0700, Stephen Boyd wrote:

> Signed-off-by: Stephen Boyd <bebarino@gmail.com>

There is no rationale given in the commit message, but I assume it is
"this spews useless crap to the terminal that interferes with the usual
test output".

However, this seems like the wrong direction. We already have a nice
test infrastructure that redirects test output away from the user and
respects "--verbose" to turn it back on. t5500 does quite a bit of work
outside the test harness, and it makes more sense to me to move it
inside. That would take care of your useless spew, _and_ it would help
detect failures in a more robust way.

In other words:

> diff --git a/t/t5500-fetch-pack.sh b/t/t5500-fetch-pack.sh
> index c450f33..ce74508 100755
> --- a/t/t5500-fetch-pack.sh
> +++ b/t/t5500-fetch-pack.sh
> @@ -97,7 +97,7 @@ pull_to_client () {
>  (
>  	mkdir client &&
>  	cd client &&
> -	git init 2>> log2.txt &&
> +	git init 2>> log2.txt > /dev/null &&
>  	git config transfer.unpacklimit 0
>  )

Why not:

  test_expect_success 'setup client repo' '
    mkdir client &&
    (cd client &&
     git init 2>>log2.txt &&
     git config transfer.unpacklimit 0
    )
  '

?

-Peff