[LTP] [PATCH 0/4] Add dependency checking

[LTP] [PATCH 0/4] Add dependency checking

[Punit Agrawal]

Hi,

While investigating failures from nightly runs, I came across some
tests that don't explicitly check for dependencies.

Add appropriate check for dependencies so the tests are reported as an
issue with the configuration rather than a test failure. These set of
patches transitions ~50-60 tests from reporting failures to a
configuration issue.

Thanks,
Punit

Punit Agrawal (4):
  controllers/cgroup_fj_common.sh: Check for dependencies during setup
  quota_remount_test01: Add explicit check for dependencies
  memcg_lib.sh: Check for killall
  controllers/pids.sh: Check for killall

 testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh | 1 +
 testcases/kernel/controllers/memcg/functional/memcg_lib.sh | 3 +++
 testcases/kernel/controllers/pids/pids.sh                  | 1 +
 testcases/kernel/fs/quota_remount/quota_remount_test01.sh  | 4 ++++
 4 files changed, 9 insertions(+)

-- 
2.17.1

[LTP] [PATCH 1/4] controllers/cgroup_fj_common.sh: Check for dependencies during setup

[Punit Agrawal]

cgroup_fj_common.sh uses 'killall' in cleanup but doesn't check for it
during setup. Add killall as an explicit dependency to bail out on
systems where it's not available.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
---
 testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh b/testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh
index 907881ca8..e163e5394 100755
--- a/testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh
+++ b/testcases/kernel/controllers/cgroup_fj/cgroup_fj_common.sh
@@ -83,6 +83,7 @@ create_subgroup()
 setup()
 {
     tst_require_root
+    tst_check_cmds killall
 
     if [ ! -f /proc/cgroups ]; then
         tst_brkm TCONF "Kernel does not support for control groups"
-- 
2.17.1

[LTP] [PATCH 2/4] quota_remount_test01: Add explicit check for dependencies

[Punit Agrawal]

quota_remount_test01 doesn't check for dependencies before
execution. Add explicit checks for quotacheck and quotaon and bail out
if they are not present.

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
---
 testcases/kernel/fs/quota_remount/quota_remount_test01.sh | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/testcases/kernel/fs/quota_remount/quota_remount_test01.sh b/testcases/kernel/fs/quota_remount/quota_remount_test01.sh
index 24f41aec1..8fa9ffc9f 100755
--- a/testcases/kernel/fs/quota_remount/quota_remount_test01.sh
+++ b/testcases/kernel/fs/quota_remount/quota_remount_test01.sh
@@ -41,6 +41,8 @@ export TCID=quota_remount_test01
 export TST_TOTAL=1
 export TST_COUNT=0
 
+. test.sh
+
 if [ -z $TMPDIR ]
 then
 	TMPDIR=/tmp
@@ -58,6 +60,8 @@ if [ ! -d /proc/sys/fs/quota ]; then
         exit 0
 fi
 
+tst_check_cmds quotacheck quotaon
+
 die()
 {
 	echo >&2 $2
-- 
2.17.1

[LTP] [PATCH 3/4] memcg_lib.sh: Check for killall

[Punit Agrawal]

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
---
 testcases/kernel/controllers/memcg/functional/memcg_lib.sh | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
index 7892f78c3..838706e1a 100755
--- a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
+++ b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
@@ -34,6 +34,9 @@ if [ $? -ne 0 ]; then
 	tst_brkm TBROK "getconf PAGESIZE failed"
 fi
 
+# Check for dependencies
+tst_check_cmds killall
+
 # Post 4.16 kernel updates stat in batch (> 32 pages) every time
 PAGESIZES=$(( $PAGESIZE * 33 ))
 
-- 
2.17.1

[LTP] [PATCH 4/4] controllers/pids.sh: Check for killall

[Punit Agrawal]

Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
---
 testcases/kernel/controllers/pids/pids.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/testcases/kernel/controllers/pids/pids.sh b/testcases/kernel/controllers/pids/pids.sh
index c5430eed7..fab33c415 100755
--- a/testcases/kernel/controllers/pids/pids.sh
+++ b/testcases/kernel/controllers/pids/pids.sh
@@ -32,6 +32,7 @@ TST_TESTFUNC=do_test
 TST_POS_ARGS=3
 TST_USAGE=usage
 TST_NEEDS_ROOT=1
+TST_NEEDS_CMDS="killall"
 
 . tst_test.sh
 
-- 
2.17.1

[LTP] [PATCH 0/4] Add dependency checking

[Jan Stancek]

----- Original Message -----
> Hi,
> 
> While investigating failures from nightly runs, I came across some
> tests that don't explicitly check for dependencies.
> 
> Add appropriate check for dependencies so the tests are reported as an
> issue with the configuration rather than a test failure. These set of
> patches transitions ~50-60 tests from reporting failures to a
> configuration issue.
> 
> Thanks,
> Punit
> 

Hi,

Pushed with the exception of patch 2.

Switching quota_remount_test01.sh away from apicmds will
likely need more changes, e.g. it should use at least 
tst_exit().

If we take the effort to port this test to shell library,
it should be "tst_test.h" one.

If you're up to rewriting test for tst_test.h, that would
be preffered. Otherwise, I rather use some shell check
and avoid including "test.h" for the sole purpose of having
tst_check_cmds.

Regards,
Jan

[LTP] [PATCH 0/4] Add dependency checking

[Punit Agrawal]

Jan Stancek <jstancek@redhat.com> writes:

> ----- Original Message -----
>> Hi,
>> 
>> While investigating failures from nightly runs, I came across some
>> tests that don't explicitly check for dependencies.
>> 
>> Add appropriate check for dependencies so the tests are reported as an
>> issue with the configuration rather than a test failure. These set of
>> patches transitions ~50-60 tests from reporting failures to a
>> configuration issue.
>> 
>> Thanks,
>> Punit
>> 
>
> Hi,
>
> Pushed with the exception of patch 2.

Thanks for merging the patches.

> Switching quota_remount_test01.sh away from apicmds will
> likely need more changes, e.g. it should use at least 
> tst_exit().
>
> If we take the effort to port this test to shell library,
> it should be "tst_test.h" one.
>
> If you're up to rewriting test for tst_test.h, that would
> be preffered. Otherwise, I rather use some shell check
> and avoid including "test.h" for the sole purpose of having
> tst_check_cmds.

Ok, makes sense to use a shell check. I'll send an updated patch.

I won't be able to get to rewriting the test anytime soon but I'd really
like to eliminate the false positives on the nightly runs so we can
start paying attention to the real failures on new development.

Regards,
Punit