From mboxrd@z Thu Jan  1 00:00:00 1970
From: Cyril Hrubis <chrubis@suse.cz>
Date: Mon, 22 Aug 2016 14:31:40 +0200
Subject: [LTP] [RFC PATCH] test.sh: add SHOULD_PASS,
	SHOULD_FAIL functions
In-Reply-To: <1471617816-15339-1-git-send-email-stanislav.kholmanskikh@oracle.com>
References: <1471617816-15339-1-git-send-email-stanislav.kholmanskikh@oracle.com>
Message-ID: <20160822123140.GA18142@rei>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it

Hi!
> Sometimes we need to execute a command and call tst_resm TPASS/TFAIL
> based on the command's exit status.
> 
> The existing ROD() function can make 99% of the job, we just
> need to let it know how the command's exit code should be
> interpreted. This patch does it and introduce a couple of new
> functions to help with the described situation.
> 
> Signed-off-by: Stanislav Kholmanskikh <stanislav.kholmanskikh@oracle.com>
> ---
> This is to help with situations like:
> 
> echo 1.0 > memory.limit_in_bytes 2> /dev/null
> if [ $? -ne 0 ]; then
>    tst_resm TPASS "return value is $?"
> else
>    tst_resm TFAIL "return value is 0"
> fi
> 
> which could be transformed to:
> 
> SHOULD_FAIL echo 1.0 \> memory.limit_in_bytes

This is quite nice. Maybe it would better be named EXPECT_FAIL and
EXPECT_PASS or something but SHOULD_ prefix is fine as well.

We should include documentation for these in this commit as well.

A few comments to the implementation below.

>  testcases/lib/test.sh |   50 +++++++++++++++++++++++++++++++++++++++++++++---
>  1 files changed, 46 insertions(+), 4 deletions(-)
> 
> diff --git a/testcases/lib/test.sh b/testcases/lib/test.sh
> index bd66109..ca2c00b 100644
> --- a/testcases/lib/test.sh
> +++ b/testcases/lib/test.sh
> @@ -220,12 +220,17 @@ ROD_SILENT()
>  	fi
>  }
>  
> -ROD()
> +ROD_DISPATCHER()
>  {
> +	local act
>  	local cmd
>  	local arg
>  	local file
>  	local flag
> +	local ret
> +
> +	act="$1"
> +	shift
>  
>  	for arg; do
>  		file="${arg#\>}"
> @@ -251,9 +256,46 @@ ROD()
>  		$@
>  	fi
>  
> -	if [ $? -ne 0 ]; then
> -		tst_brkm TBROK "$@ failed"
> -	fi
> +	ret=$?

Why don't we just return from this function here and let the caller
examine the $?

> +	case "$act" in
> +		0) # break on failure
> +		if [ $ret -ne 0 ]; then
> +			tst_brkm TBROK "$@ failed"
> +		fi;;
> +
> +		1) # the command should pass
> +		if [ $ret -eq 0 ]; then
> +			tst_resm TPASS "$@ passed as expected"
> +		else
> +			tst_resm TFAIL "$@ failed unexpectedly"
> +		fi;;
> +
> +		2) # the command should fail
> +		if [ $ret -ne 0 ]; then
> +			tst_resm TPASS "$@ failed as expected"
> +		else
> +			tst_resm TFAIL "$@ passed unexpectedly"
> +		fi;;
> +
> +		*) tst_brkm TBROK "unknown action '$act'";;
> +	esac
> +}
> +
> +ROD()
> +{
> +	ROD_DISPATCHER 0 $@
> +}
> +
> +SHOULD_PASS()
> +{
> +	ROD_DISPATCHER 1 $@
> +}
> +
> +SHOULD_FAIL()
> +{
> +	# redirect stderr since we expect the command to fail
> +	ROD_DISPATCHER 2 $@ 2> /dev/null
>  }

I think that we should quote the $@ in these functions.

Try for yourself:

8<------------------------------------

#!/bin/sh

d()
{
        echo "-------------"
        for i; do
                echo $i
        done
        echo "-------------"
}

dd()
{
        d "$@"
}


d a b c
dd a b c
d "a b c"
dd "a b c"

8<------------------------------------

If you unquote the $@ in dd() the "a b c" in the last call would be separated
on spaces.

-- 
Cyril Hrubis
chrubis@suse.cz