diff mbox series

selftests: ksft: Track skipped tests when finishing the test suite

Message ID 20240722154319.619944-1-laura.nao@collabora.com (mailing list archive)
State Accepted
Commit 170c966cbe274e664288cfc12ee919d5e706dc50
Headers show
Series selftests: ksft: Track skipped tests when finishing the test suite | expand

Commit Message

Laura Nao July 22, 2024, 3:43 p.m. UTC
Consider skipped tests in addition to passed tests when evaluating the
overall result of the test suite in the finished() helper.

Signed-off-by: Laura Nao <laura.nao@collabora.com>
---
 tools/testing/selftests/kselftest/ksft.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Shuah Khan July 22, 2024, 5:32 p.m. UTC | #1
On 7/22/24 09:43, Laura Nao wrote:
> Consider skipped tests in addition to passed tests when evaluating the
> overall result of the test suite in the finished() helper.
> 
> Signed-off-by: Laura Nao <laura.nao@collabora.com>
> ---
>   tools/testing/selftests/kselftest/ksft.py | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
> index cd89fb2bc10e..bf215790a89d 100644
> --- a/tools/testing/selftests/kselftest/ksft.py
> +++ b/tools/testing/selftests/kselftest/ksft.py
> @@ -70,7 +70,7 @@ def test_result(condition, description=""):
>   
>   
>   def finished():
> -    if ksft_cnt["pass"] == ksft_num_tests:
> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:

Please don't. Counting skips in pass or fail isn't accurate
reporting. skips need to be reported as skips.

thanks,
-- Shuah
Shuah Khan July 22, 2024, 6:42 p.m. UTC | #2
On 7/22/24 11:32, Shuah Khan wrote:
> On 7/22/24 09:43, Laura Nao wrote:
>> Consider skipped tests in addition to passed tests when evaluating the
>> overall result of the test suite in the finished() helper.
>>
>> Signed-off-by: Laura Nao <laura.nao@collabora.com>
>> ---
>>   tools/testing/selftests/kselftest/ksft.py | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
>> index cd89fb2bc10e..bf215790a89d 100644
>> --- a/tools/testing/selftests/kselftest/ksft.py
>> +++ b/tools/testing/selftests/kselftest/ksft.py
>> @@ -70,7 +70,7 @@ def test_result(condition, description=""):
>>   def finished():
>> -    if ksft_cnt["pass"] == ksft_num_tests:
>> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
> 
> Please don't. Counting skips in pass or fail isn't accurate
> reporting. skips need to be reported as skips.
> 

More on this since I keep seeing patches like this one that
make the reporting confusing.

There is a reason why you don't want to mark a test passed
when there are several skips. Skips are an indication that
there are several tests and/or test cases that couldn't not
be run because of unmet dependencies. This condition needs
to be investigated to see if there are any config options
that could be enabled to get a better coverage.

Including skips to determine pass gives a false sense security
that all is well when it isn't

thanks,
-- Shuah
Nícolas F. R. A. Prado July 22, 2024, 6:51 p.m. UTC | #3
On Mon, Jul 22, 2024 at 11:32:35AM -0600, Shuah Khan wrote:
> On 7/22/24 09:43, Laura Nao wrote:
> > Consider skipped tests in addition to passed tests when evaluating the
> > overall result of the test suite in the finished() helper.
> > 
> > Signed-off-by: Laura Nao <laura.nao@collabora.com>
> > ---
> >   tools/testing/selftests/kselftest/ksft.py | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
> > index cd89fb2bc10e..bf215790a89d 100644
> > --- a/tools/testing/selftests/kselftest/ksft.py
> > +++ b/tools/testing/selftests/kselftest/ksft.py
> > @@ -70,7 +70,7 @@ def test_result(condition, description=""):
> >   def finished():
> > -    if ksft_cnt["pass"] == ksft_num_tests:
> > +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
> 
> Please don't. Counting skips in pass or fail isn't accurate
> reporting. skips need to be reported as skips.

Hi Shuah,

this won't change the skip count, just allow a test suite that has a mix of pass
and skip results to exit with code 0. That's the same behavior as the C
ksft_finished() helper in kselftest.h:

#define ksft_finished()			\
	ksft_exit(ksft_plan ==		\
		  ksft_cnt.ksft_pass +	\
		  ksft_cnt.ksft_xfail +	\
		  ksft_cnt.ksft_xskip)

It was my oversight to not do the same in the python helper.

Laura,

I consider this fixing an incorrect behavior, so I'd add this tag:

Fixes: dacf1d7a78bf ("kselftest: Add test to verify probe of devices from discoverable buses")

I think the message is good as is, but maybe it could have mentioned that this
matches the behavior of the C helper, just to make the point above clearer.

Reviewed-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>

And just a note for the maintainers, this patch depends on "kselftest: Move ksft
helper module to common directory"
https://lore.kernel.org/all/20240705-dev-err-log-selftest-v2-2-163b9cd7b3c1@collabora.com/
which was picked through the usb tree but is queued for 6.11-rc1.

Thanks,
Nícolas
Laura Nao July 23, 2024, 2:06 p.m. UTC | #4
On 7/22/24 20:51, Nícolas F. R. A. Prado wrote:
> On Mon, Jul 22, 2024 at 11:32:35AM -0600, Shuah Khan wrote:
>> On 7/22/24 09:43, Laura Nao wrote:
>>> Consider skipped tests in addition to passed tests when evaluating the
>>> overall result of the test suite in the finished() helper.
>>>
>>> Signed-off-by: Laura Nao <laura.nao@collabora.com>
>>> ---
>>>    tools/testing/selftests/kselftest/ksft.py | 2 +-
>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
>>> index cd89fb2bc10e..bf215790a89d 100644
>>> --- a/tools/testing/selftests/kselftest/ksft.py
>>> +++ b/tools/testing/selftests/kselftest/ksft.py
>>> @@ -70,7 +70,7 @@ def test_result(condition, description=""):
>>>    def finished():
>>> -    if ksft_cnt["pass"] == ksft_num_tests:
>>> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
>>
>> Please don't. Counting skips in pass or fail isn't accurate
>> reporting. skips need to be reported as skips.
> 
> Hi Shuah,
> 
> this won't change the skip count, just allow a test suite that has a mix of pass
> and skip results to exit with code 0. That's the same behavior as the C
> ksft_finished() helper in kselftest.h:
> 
> #define ksft_finished()			\
> 	ksft_exit(ksft_plan ==		\
> 		  ksft_cnt.ksft_pass +	\
> 		  ksft_cnt.ksft_xfail +	\
> 		  ksft_cnt.ksft_xskip)
> 

Correct, this patch aligns the behavior of the python helper with both 
the existing C and bash counterparts (ksft_finished() and 
ktap_finished() respectively).

> It was my oversight to not do the same in the python helper.
> 
> Laura,
> 
> I consider this fixing an incorrect behavior, so I'd add this tag:
> 
> Fixes: dacf1d7a78bf ("kselftest: Add test to verify probe of devices from discoverable buses")
> 
> I think the message is good as is, but maybe it could have mentioned that this
> matches the behavior of the C helper, just to make the point above clearer.
> 
> Reviewed-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>
> 
> And just a note for the maintainers, this patch depends on "kselftest: Move ksft
> helper module to common directory"
> https://lore.kernel.org/all/20240705-dev-err-log-selftest-v2-2-163b9cd7b3c1@collabora.com/
> which was picked through the usb tree but is queued for 6.11-rc1.
> 

Right, thanks!

Shuah, please let me know if you have any other concerns with this patch.

Best,

Laura
Shuah Khan July 23, 2024, 4:17 p.m. UTC | #5
On 7/22/24 09:43, Laura Nao wrote:
> Consider skipped tests in addition to passed tests when evaluating the
> overall result of the test suite in the finished() helper.
> 
> Signed-off-by: Laura Nao <laura.nao@collabora.com>
> ---
>   tools/testing/selftests/kselftest/ksft.py | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
> index cd89fb2bc10e..bf215790a89d 100644
> --- a/tools/testing/selftests/kselftest/ksft.py
> +++ b/tools/testing/selftests/kselftest/ksft.py
> @@ -70,7 +70,7 @@ def test_result(condition, description=""):
>   
>   
>   def finished():
> -    if ksft_cnt["pass"] == ksft_num_tests:
> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
>           exit_code = KSFT_PASS

Laura and Nícolas,

I saw both your emails explaining how this fixes the problem in
a previous patch.

However looks like you haven't see my response about the implications
of the exit_code = KSFT_PASS when tests are skipped.

if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
>           exit_code = KSFT_PASS

Let me reiterate in case you missed it:

There is a reason why you don't want to mark all tests passed
when there are several skips.Skips are an indication that
there are several tests and/or test cases that couldn't not
be run because of unmet dependencies. This condition needs
to be investigated to see if there are any config options
that could be enabled to get a better coverage.

Including skips to determine pass gives a false sense security
that all is well when it isn't.

So it is incorrect to set the exit code to KSFT_PASS when there
are skipped tests.

+    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:


>       else:
>           exit_code = KSFT_FAIL

The logic here seems to not take into account when you have a
conditions where you have a mixed results of passed tests,
skipped tests, and failed tests.

Please revisit and figure out how to address this and report
the status correctly.

thanks,
-- Shuah
Laura Nao July 29, 2024, 2:52 p.m. UTC | #6
Hi Shuah,

On 7/23/24 18:17, Shuah Khan wrote:
> On 7/22/24 09:43, Laura Nao wrote:
>> Consider skipped tests in addition to passed tests when evaluating the
>> overall result of the test suite in the finished() helper.
>>
>> Signed-off-by: Laura Nao <laura.nao@collabora.com>
>> ---
>>   tools/testing/selftests/kselftest/ksft.py | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/kselftest/ksft.py 
>> b/tools/testing/selftests/kselftest/ksft.py
>> index cd89fb2bc10e..bf215790a89d 100644
>> --- a/tools/testing/selftests/kselftest/ksft.py
>> +++ b/tools/testing/selftests/kselftest/ksft.py
>> @@ -70,7 +70,7 @@ def test_result(condition, description=""):
>>   def finished():
>> -    if ksft_cnt["pass"] == ksft_num_tests:
>> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
>>           exit_code = KSFT_PASS
> 
> Laura and Nícolas,
> 
> I saw both your emails explaining how this fixes the problem in
> a previous patch.
> 
> However looks like you haven't see my response about the implications
> of the exit_code = KSFT_PASS when tests are skipped.
> 
> if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
>>           exit_code = KSFT_PASS
> 
> Let me reiterate in case you missed it:
> 
> There is a reason why you don't want to mark all tests passed
> when there are several skips.Skips are an indication that
> there are several tests and/or test cases that couldn't not
> be run because of unmet dependencies. This condition needs
> to be investigated to see if there are any config options
> that could be enabled to get a better coverage.
> 
> Including skips to determine pass gives a false sense security
> that all is well when it isn't.
> 
> So it is incorrect to set the exit code to KSFT_PASS when there
> are skipped tests.

Just to be clear, the logic in ksft_finished() in kselftest.h (the C
helper) is to exit with KSFT_PASS when there are only passed and skipped 
tests. You're suggesting we change it to exit with KSFT_FAIL in that 
case?

Under this new logic, the runner would effectively treat skips as 
failures, impacting existing kselftests.

> +    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
> 
> 
>>       else:
>>           exit_code = KSFT_FAIL
> 
> The logic here seems to not take into account when you have a
> conditions where you have a mixed results of passed tests,
> skipped tests, and failed tests.
>
> Please revisit and figure out how to address this and report
> the status correctly.

The logic ensures that if there’s a mix of passed, skipped, and failed 
tests, the exit code returned is KSFT_FAIL. I believe that if there is at 
least one failed test, the overall test should be reported as failed, 
which is what happens in this case.

Thanks,

Laura
Shuah Khan July 29, 2024, 8:06 p.m. UTC | #7
On 7/29/24 08:52, Laura Nao wrote:
> Hi Shuah,
> 
> On 7/23/24 18:17, Shuah Khan wrote:
>> On 7/22/24 09:43, Laura Nao wrote:
>>> Consider skipped tests in addition to passed tests when evaluating the
>>> overall result of the test suite in the finished() helper.
>>>

I am finally with you now. Can you please more information in your
short log and commit log.

Isn't this a bug fix? Current logic before this change will report
tests failed if there are any skipped tests?

Can you send v2 calling it a fix and explain the problem clearly.
  
This isn't problem in this patch, but I am concerned about how
simply calling tests passed without calling out skipped tests.

This problem could be solved by printing a message at the end of tests
for non-zero skipped test conditions to say the coverage could be
increased by enabling config options.

thanks,
-- Shuah
Laura Nao July 30, 2024, 10:35 a.m. UTC | #8
On 7/29/24 22:06, Shuah Khan wrote:
> On 7/29/24 08:52, Laura Nao wrote:
>> Hi Shuah,
>>
>> On 7/23/24 18:17, Shuah Khan wrote:
>>> On 7/22/24 09:43, Laura Nao wrote:
>>>> Consider skipped tests in addition to passed tests when evaluating the
>>>> overall result of the test suite in the finished() helper.
>>>>
> 
> I am finally with you now. Can you please more information in your
> short log and commit log.
> 
> Isn't this a bug fix? Current logic before this change will report
> tests failed if there are any skipped tests?
> 
> Can you send v2 calling it a fix and explain the problem clearly.
>

v2 sent: https://lore.kernel.org/all/20240730102928.33182-1-laura.nao@collabora.com/
 
> This isn't problem in this patch, but I am concerned about how
> simply calling tests passed without calling out skipped tests.
> 
> This problem could be solved by printing a message at the end of tests
> for non-zero skipped test conditions to say the coverage could be
> increased by enabling config options.
>

Right. We can send a separate patch to address this.

Thanks!

Laura
Shuah Khan July 31, 2024, 11:04 p.m. UTC | #9
On 7/30/24 04:35, Laura Nao wrote:
> On 7/29/24 22:06, Shuah Khan wrote:
>> On 7/29/24 08:52, Laura Nao wrote:
>>> Hi Shuah,
>>>
>>> On 7/23/24 18:17, Shuah Khan wrote:
>>>> On 7/22/24 09:43, Laura Nao wrote:
>>>>> Consider skipped tests in addition to passed tests when evaluating the
>>>>> overall result of the test suite in the finished() helper.
>>>>>
>>
>> I am finally with you now. Can you please more information in your
>> short log and commit log.
>>
>> Isn't this a bug fix? Current logic before this change will report
>> tests failed if there are any skipped tests?
>>
>> Can you send v2 calling it a fix and explain the problem clearly.
>>
> 
> v2 sent: https://lore.kernel.org/all/20240730102928.33182-1-laura.nao@collabora.com/
>   
>> This isn't problem in this patch, but I am concerned about how
>> simply calling tests passed without calling out skipped tests.
>>
>> This problem could be solved by printing a message at the end of tests
>> for non-zero skipped test conditions to say the coverage could be
>> increased by enabling config options.
>>
> 
> Right. We can send a separate patch to address this.

Yes please. Thank you.

thanks,
-- Shuah
diff mbox series

Patch

diff --git a/tools/testing/selftests/kselftest/ksft.py b/tools/testing/selftests/kselftest/ksft.py
index cd89fb2bc10e..bf215790a89d 100644
--- a/tools/testing/selftests/kselftest/ksft.py
+++ b/tools/testing/selftests/kselftest/ksft.py
@@ -70,7 +70,7 @@  def test_result(condition, description=""):
 
 
 def finished():
-    if ksft_cnt["pass"] == ksft_num_tests:
+    if ksft_cnt["pass"] + ksft_cnt["skip"] == ksft_num_tests:
         exit_code = KSFT_PASS
     else:
         exit_code = KSFT_FAIL