| Message ID | 20201217144424.19414-2-pvorel@suse.cz |
|---|---|
| State | Accepted |
| Headers | show |
| Series | shell API setup/cleanup checks | expand |
Hi! > It does not make sense to run the test cleanup function when the setup > function has been run. Did you mean "setup function hasn't been run." here? As it is this sentence does not make any sense to me. > And at least some network tests expect setup has been run before running > cleanup (e.g. tcp_fastopen_run.sh). > > When shell API was introduced, cleanup function was run only if 1) setup > function was defined 2) and also run. That was inconsistent from C API, > thus e7dc14caa run it always. > > But shell API is different from C API: tst_brk can be called from > tst_test.sh (or other library which is run before tst_run, e.g. > tst_net.sh). That was probably the reason, why detection via > $TST_SETUP_STARTED was introduced in initial shell API. > > NOTE: using type is better than grep $TST_TEST_PATH, because cleanup > function can be in other library sourced by the test. > > Fixes: e7dc14caa ("tst_test.sh: Run cleanup even setup is not defined") This still possibly break tests that would call tst_brk() in the middle of setup and expect the cleanup() to be executed, right? I guess that we would need TST_DO_CLEANUP that would be set in both cases i.e. before we run setup and also before we execute the tests. What about this: diff --git a/testcases/lib/tst_test.sh b/testcases/lib/tst_test.sh index 2417da140..94d95df6f 100644 --- a/testcases/lib/tst_test.sh +++ b/testcases/lib/tst_test.sh @@ -28,7 +28,7 @@ _tst_do_exit() local ret=0 TST_DO_EXIT=1 - if [ -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then + if [ -n $TST_DO_CLEANUP -a -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then $TST_CLEANUP fi @@ -582,6 +582,7 @@ tst_run() [ -n "$TST_NEEDS_MODULE" ] && tst_require_module "$TST_NEEDS_MODULE" if [ -n "$TST_SETUP" ]; then + TST_DO_CLEANUP=1 $TST_SETUP fi @@ -608,6 +609,7 @@ _tst_run_tests() local _tst_data="$1" local _tst_i + TST_DO_CLEANUP=1 for _tst_i in $(seq ${TST_CNT:-1}); do if type ${TST_TESTFUNC}1 > /dev/null 2>&1; then _tst_run_test "$TST_TESTFUNC$_tst_i" $_tst_i "$_tst_data"
Hi Cyril, > Hi! > > It does not make sense to run the test cleanup function when the setup > > function has been run. > Did you mean "setup function hasn't been run." here? > As it is this sentence does not make any sense to me. Yes, thanks! > > And at least some network tests expect setup has been run before running > > cleanup (e.g. tcp_fastopen_run.sh). > > When shell API was introduced, cleanup function was run only if 1) setup > > function was defined 2) and also run. That was inconsistent from C API, > > thus e7dc14caa run it always. > > But shell API is different from C API: tst_brk can be called from > > tst_test.sh (or other library which is run before tst_run, e.g. > > tst_net.sh). That was probably the reason, why detection via > > $TST_SETUP_STARTED was introduced in initial shell API. > > NOTE: using type is better than grep $TST_TEST_PATH, because cleanup > > function can be in other library sourced by the test. > > Fixes: e7dc14caa ("tst_test.sh: Run cleanup even setup is not defined") > This still possibly break tests that would call tst_brk() in the middle > of setup and expect the cleanup() to be executed, right? > I guess that we would need TST_DO_CLEANUP that would be set in both > cases i.e. before we run setup and also before we execute the tests. > What about this: Good catch, agree. I'm going to merge with one runtime fix: > - if [ -n $TST_DO_CLEANUP -a -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then > + if [ -n "$TST_DO_CLEANUP" -a -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then Quotes are required otherwise we can get error: /opt/ltp/testcases/bin/tst_test.sh: line 31: [: too many arguments Kind regards, Petr > diff --git a/testcases/lib/tst_test.sh b/testcases/lib/tst_test.sh > index 2417da140..94d95df6f 100644 > --- a/testcases/lib/tst_test.sh > +++ b/testcases/lib/tst_test.sh > @@ -28,7 +28,7 @@ _tst_do_exit() > local ret=0 > TST_DO_EXIT=1 > - if [ -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then > + if [ -n $TST_DO_CLEANUP -a -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then > $TST_CLEANUP > fi > @@ -582,6 +582,7 @@ tst_run() > [ -n "$TST_NEEDS_MODULE" ] && tst_require_module "$TST_NEEDS_MODULE" > if [ -n "$TST_SETUP" ]; then > + TST_DO_CLEANUP=1 > $TST_SETUP > fi > @@ -608,6 +609,7 @@ _tst_run_tests() > local _tst_data="$1" > local _tst_i > + TST_DO_CLEANUP=1 > for _tst_i in $(seq ${TST_CNT:-1}); do > if type ${TST_TESTFUNC}1 > /dev/null 2>&1; then > _tst_run_test "$TST_TESTFUNC$_tst_i" $_tst_i "$_tst_data"
Hi Cyril, ... > This still possibly break tests that would call tst_brk() in the middle > of setup and expect the cleanup() to be executed, right? > I guess that we would need TST_DO_CLEANUP that would be set in both > cases i.e. before we run setup and also before we execute the tests. > What about this: Whole patchset merged with your credit. Thanks! Kind regards, Petr
diff --git a/doc/test-writing-guidelines.txt b/doc/test-writing-guidelines.txt index 99fb34628..f3a55cf26 100644 --- a/doc/test-writing-guidelines.txt +++ b/doc/test-writing-guidelines.txt @@ -2251,11 +2251,15 @@ able to clean up correctly even in this situation. The easiest solution for this is to keep track of what was initialized and act accordingly in the cleanup. -WARNING: Similar to the C library, calling tst_brk() in the $TST_CLEANUP does +WARNING: Similar to the C library, calling 'tst_brk' in the $TST_CLEANUP does not exit the test and 'TBROK' is converted to 'TWARN'. -Notice also the 'tst_run' function called at the end of the test that actually -starts the test. +Notice also the 'tst_run' shell API function called at the end of the test that +actually starts the test. + +WARNING: cleanup function is called only after 'tst_run' has been started. +Calling 'tst_brk' in shell libraries, e.g. 'tst_test.sh' or 'tst_net.sh' does +not trigger calling it. [source,sh] ------------------------------------------------------------------------------- diff --git a/testcases/lib/tst_test.sh b/testcases/lib/tst_test.sh index 2417da140..c205bc91b 100644 --- a/testcases/lib/tst_test.sh +++ b/testcases/lib/tst_test.sh @@ -28,7 +28,7 @@ _tst_do_exit() local ret=0 TST_DO_EXIT=1 - if [ -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then + if [ -n "$TST_RUN_STARTED" -a -n "$TST_CLEANUP" -a -z "$TST_NO_CLEANUP" ]; then $TST_CLEANUP fi @@ -599,7 +599,6 @@ tst_run() fi TST_ITERATIONS=$((TST_ITERATIONS-1)) done - _tst_do_exit } @@ -608,6 +607,7 @@ _tst_run_tests() local _tst_data="$1" local _tst_i + TST_RUN_STARTED=1 for _tst_i in $(seq ${TST_CNT:-1}); do if type ${TST_TESTFUNC}1 > /dev/null 2>&1; then _tst_run_test "$TST_TESTFUNC$_tst_i" $_tst_i "$_tst_data"
It does not make sense to run the test cleanup function when the setup function has been run. And at least some network tests expect setup has been run before running cleanup (e.g. tcp_fastopen_run.sh). When shell API was introduced, cleanup function was run only if 1) setup function was defined 2) and also run. That was inconsistent from C API, thus e7dc14caa run it always. But shell API is different from C API: tst_brk can be called from tst_test.sh (or other library which is run before tst_run, e.g. tst_net.sh). That was probably the reason, why detection via $TST_SETUP_STARTED was introduced in initial shell API. NOTE: using type is better than grep $TST_TEST_PATH, because cleanup function can be in other library sourced by the test. Fixes: e7dc14caa ("tst_test.sh: Run cleanup even setup is not defined") Signed-off-by: Petr Vorel <pvorel@suse.cz> --- doc/test-writing-guidelines.txt | 10 +++++++--- testcases/lib/tst_test.sh | 4 ++-- 2 files changed, 9 insertions(+), 5 deletions(-)