[OE-core] [PATCH 1/1] oeqa/runtime/cases/rpm.py: change to use base-passwd-doc

[Alexander Kanavin]

Right, I see. Thanks for explaining, the patch is fine then.

Alex

2018-06-14 12:31 GMT+03:00 ChenQi <Qi.Chen at windriver.com>:
> Hi Alex,
>
> In short, it's about setUpClass.
> See details below.
>
> In fact, if rpm is not installed, these test cases have been skipped.
>
> See codes below from rpm.py:
>
>     @OETestID(960)
>     @OEHasPackage(['rpm'])
>     @OETestDepends(['ssh.SSHTest.test_ssh'])
>     def test_rpm_help(self):
>         status, output = self.target.run('rpm --help')
>         msg = 'status and output: %s and %s' % (status, output)
>         self.assertEqual(status, 0, msg=msg)
>
>     @OETestID(192)
>     @OETestDepends(['rpm.RpmBasicTest.test_rpm_help'])
>     def test_rpm_install(self):
>         status, output = self.target.run('rpm -ivh
> /tmp/base-passwd-doc.rpm')
>         msg = 'Failed to install base-passwd-doc package: %s' % output
>         self.assertEqual(status, 0, msg=msg)
>
> You can see that if 'rpm' is not installed on target, the test cases are
> skipped.
> The '@skipIfNotFeature' decorator you suggested will achieve the same effect
> that the above logic has already achieved.
>
> The key problem here is that the copying action happens in setUpClass().
> Even if the testcase is determined to be skipped via the decorator logic,
> this setUpClass function will still be called.
>
> For the current OEQA's runtime testing design, we need to ensure setUpClass
> do not generate 'error'.
>
> For the problem I described in this email, the simplest approach I see here
> is to change rpm-doc to base-passwd-doc.
> Another apporach is to check whether 'rpm' is installed in setUpClass()
> function and if not skip. But this is more complex and kind of weird as
> we've checked it via @HasPackage and the test cases dependencies are
> correct.
>
> You can do a quick test with the following configuration to reproduce the
> problem.
> local.conf:
> IMAGE_INSTALL_append = " openssh"
> TEST_IMAGE = "1"
> TEST_SUITES = "auto"
>
> bitbake core-image-minimal
>
> Best Regards,
> Chen Qi
>
>
>
> On 06/14/2018 04:55 PM, Alexander Kanavin wrote:
>>
>> Hello Chen,
>>
>> But the test still attempts to run rpm on the target. I can't
>> understand how target rpm can be available on the image, but at the
>> same time target rpm-doc package is not available in rpm deploy dir.
>>
>> The issue is probably that the test should simply be skipped (in the
>> same way that dnf testcases are skipped - please refer to them), if
>> there is no package management on target? Like this:
>>
>>      @skipIfNotFeature('package-management',
>>                        'Test requires package-management to be in
>> IMAGE_FEATURES')
>>
>>
>> Alex
>>
>> 2018-06-14 10:29 GMT+03:00 ChenQi <Qi.Chen at windriver.com>:
>>>
>>> On 06/14/2018 02:57 PM, Alexander Kanavin wrote:
>>>>
>>>> 2018-06-14 7:52 GMT+03:00 Chen Qi <Qi.Chen at windriver.com>:
>>>>>
>>>>> The test cases assume that rpm-doc is built out, as it says it its
>>>>> comment. This is not always true. And it sometimes results in
>>>>> following error.
>>>>>
>>>>>     |     cls.tc.target.copyTo(test_file, dst)
>>>>>     | UnboundLocalError: local variable 'test_file' referenced before
>>>>> assignment
>>>>
>>>> Wait, you need to explain this further. What is the situation when it
>>>> isn't true?
>>>>
>>>> Alex
>>>>
>>> Hi Alex,
>>>
>>> For any image that does not have rpm installed, rpm package is not likely
>>> to
>>> be built out.
>>> Having PACKAGE_CLASSES set to "package_rpm" will only ensure rpm-native
>>> to
>>> be built out.
>>>
>>> Best Regards,
>>> Chen Qi
>
>
>