[Cluster-devel] [PATCH] gfs2-utils tests: Add a script to exercise the utils

[Andrew Price <anprice@redhat.com>]

On 14/12/12 13:33, Steven Whitehouse wrote:
> Hi,
>
> On Fri, 2012-12-14 at 13:21 +0000, Andrew Price wrote:
>> Add a test script to make it easy to run gfs2 utils with various options
>> and check their exit codes in sequence. The script is plugged into the
>> test suite and is run with 'make check'.
>>
>> Signed-off-by: Andrew Price <anprice@redhat.com>
>> ---
>>   tests/Makefile.am   |  4 +++-
>>   tests/tool_tests.sh | 62 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>>   2 files changed, 65 insertions(+), 1 deletion(-)
>>   create mode 100755 tests/tool_tests.sh
>>
>> diff --git a/tests/Makefile.am b/tests/Makefile.am
>> index 71c1e08..d8aa8f2 100644
>> --- a/tests/Makefile.am
>> +++ b/tests/Makefile.am
>> @@ -1,4 +1,6 @@
>> -TESTS			= check_libgfs2
>> +TESTS_ENVIRONMENT	= TOPBUILDDIR=$(top_builddir)
>> +TESTS			= check_libgfs2 tool_tests.sh
>> +EXTRA_DIST		= tool_tests.sh
>>   check_PROGRAMS		= check_libgfs2
>>   check_libgfs2_SOURCES	= check_meta.c \
>>   			  $(top_srcdir)/gfs2/libgfs2/libgfs2.h
>> diff --git a/tests/tool_tests.sh b/tests/tool_tests.sh
>> new file mode 100755
>> index 0000000..791b071
>> --- /dev/null
>> +++ b/tests/tool_tests.sh
>> @@ -0,0 +1,62 @@
>> +#!/bin/sh
>> +
>> +# This script runs gfs2 utils with various options, checking exit codes against
>> +# expected values. If any test fails to exit with an expected code, the exit code
>> +# of the whole script will be non-zero but the tests will continue to be run. The
>> +# sparse file which is used as the target of the tests can be configured by
>> +# setting the environment variables TEST_TARGET (the filename) and TEST_TARGET_SZ
>> +# (its apparent size in gigabytes). Defaults to "test_sparse" and 10GB.
>> +
>> +MKFS="${TOPBUILDDIR}/gfs2/mkfs/mkfs.gfs2 -qO"
>> +FSCK="${TOPBUILDDIR}/gfs2/fsck/fsck.gfs2 -qn"
>> +
>> +# Name of the sparse file we'll use for testing
>> +TEST_TARGET=${TEST_TARGET:-test_sparse}
>> +# Size, in GB, of the sparse file we'll create to run the tests
>> +TEST_TARGET_SZ=${TEST_TARGET_SZ:-10}
>> +[ $TEST_TARGET_SZ -gt 0 ] || { echo "Target size (in GB) must be greater than 0" >&2; exit 1; }
>> +# Overall success (so we can keep going if one test fails)
>> +TEST_RET=0
>> +
>> +fn_test()
>> +{
>> +	local expected="$1"
>> +	local cmd="$2"
>> +	echo -n "Running '$cmd' - (Exp: $expected Got: "
>> +	$cmd &> /dev/null;
>> +	local ret=$?
>> +	echo -n "$ret) "
>> +	if [ "$ret" != "$expected" ];
>> +	then
>> +		echo "FAIL"
>> +		TEST_RET=1
>> +		TEST_GRP_RET=1
>> +	else
>> +		echo "PASS"
>> +	fi
>> +}
>> +
>> +fn_rm_target()
>> +{
>> +	fn_test 0 "rm -f $TEST_TARGET"
>> +}
>> +
>> +fn_recreate_target()
>> +{
>> +	fn_rm_target
>> +	fn_test 0 "dd if=/dev/null of=$TEST_TARGET bs=1 count=0 seek=${TEST_TARGET_SZ}G"
>
> This doesn't look very sparse, shouldn't we be creating it with truncate
> rather than dd?

Looks sparse to me:

$ dd if=/dev/null of=test_sparse bs=1 count=0 seek=10G
0+0 records in
0+0 records out
0 bytes (0 B) copied, 1.4466e-05 s, 0.0 kB/s
$ du test_sparse
0	test_sparse
$ du --apparent-size test_sparse
10485760	test_sparse

Truncate would work too but unless there's a hidden problem with the 
above I'll need more convincing to change it :)

Andy

>
> Steve.
>
>
>> +}
>> +
>> +
>> +# Tests start here
>> +fn_recreate_target
>> +fn_test 0 "$MKFS -p lock_nolock $TEST_TARGET"
>> +fn_test 0 "$MKFS -p lock_dlm -t foo:bar $TEST_TARGET"
>> +fn_test 255 "$MKFS -p badprotocol $TEST_TARGET"
>> +fn_test 0 "$FSCK $TEST_TARGET"
>> +
>> +# Tests end here
>> +
>> +# Clean up
>> +fn_test 0 "rm -f $TEST_TARGET"
>> +exit $TEST_RET
>
>