Re: media/master v4l2-compliance on vivid: 236 tests, 0 regressions (media_v5.1-2-16-gfc8670d1f72b)

[Hans Verkuil <hverkuil@xs4all.nl>]

On 5/24/19 10:38 AM, Guillaume Tucker wrote:
> Hi Hans,
> 
> On 16/05/2019 07:41, Hans Verkuil wrote:
>> Hi Guillaume,
>>
>> I have a few questions/suggestions:
> 
> Thanks for the feedback!  It's good to start seeing these reports
> on the linux-media mailing list.  And sorry for the slow reply, I
> was away.
> 
>> On 5/15/19 9:04 PM, kernelci.org bot wrote:
>>> media/master v4l2-compliance on vivid: 236 tests, 0 regressions (media_v5.1-2-16-gfc8670d1f72b)
>>>
>>> Test results summary
>>> --------------------
>>>
>>> V4L2 Compliance on the vivid driver.
>>>
>>> This test ran "v4l2-compliance -s" from v4l-utils:
>>>
>>>     https://www.linuxtv.org/wiki/index.php/V4l2-utils
>>
>> I'd just link directly to the git repo instead of the wiki: https://git.linuxtv.org/v4l-utils.git
> 
> Sure, I thought this had been agreed before but it's easy to
> change.
> 
>> You should add the v4l-utils commit that's used to compile v4l2-compliance.
>> That's important information to have. I assume that this test always uses the
>> latest version of v4l-utils?
> 
> This information is in the detailed results, but the detailed
> results are only shown when there are some failures.  So we'll
> rework that a bit.
> 
> For example, from the v4l2-compliance-uvc report:
> 
> 
> Test failures
> ------------- 
> 1  | rk3399-gru-kevin       | arm64 |  52 total:  43 PASS   9 FAIL   0 SKIP
> 
>   Config:      defconfig
>   Compiler:    gcc-8 (aarch64-linux-gnu-gcc (Debian 8.3.0-2) 8.3.0)
>   Lab Name:    lab-collabora
>   Plain log:   https://storage.kernelci.org//media/master/media_v5.1-2-16-gfc8670d1f72b/arm64/defconfig/gcc-8/lab-collabora/v4l2-compliance-uvc-rk3399-gru-kevin.txt
>   HTML log:    https://storage.kernelci.org//media/master/media_v5.1-2-16-gfc8670d1f72b/arm64/defconfig/gcc-8/lab-collabora/v4l2-compliance-uvc-rk3399-gru-kevin.html
>   Rootfs:      http://storage.kernelci.org/images/rootfs/debian/stretch-v4l2/20190510.0/arm64/rootfs.cpio.gz
>   Test Git:    git://linuxtv.org/v4l-utils.git
>   Test Commit: 0d61ddede7d340ffa1c75a2882e30c455ef3d8b8
> 
> 
> The git repo and commit hash here show you which version of
> vl4-utils was used.
> 
> 
> At the moment, the v4l2-compliance is part of a rootfs which gets
> updated each time kernelci.org production code gets updated,
> which is typically once a week.  This can be improved to have the
> rootfs updates independent from the rest, then we could trigger
> rebuilds every time v4l-utils changes, but there are a few things
> to take into consideration before we can do this safely.

I don't think it has to be updated every time v4l-utils changes, at least
for now, as long as it is clear which v4l-utils version is used.

> 
>>> See each detailed section in the report below to find out the git URL and
>>> particular revision that was used to build the test binaries.
>>>
>>>
>>>   Tree:    media
>>>   Branch:  master
>>>   Kernel:  media_v5.1-2-16-gfc8670d1f72b
>>
>> I assume this is the version of the host kernel, right? Perhaps calling this
>> "Host Kernel:" would be less ambiguous.
> 
> I have to say I fail to see any ambiguity here: KernelCI is about
> testing kernels, and this tells you the kernel revision under
> test.  Calling it "host" kernel might actually be confusing when
> running with QEMU as people may think it's the version on the
> host server running the test.
> 
>>>   URL:     https://git.linuxtv.org/media_tree.git
>>>   Commit:  fc8670d1f72b746ff3a5fe441f1fca4c4dba0e6f

What confuses me is that the kernel above says: media_v5.1-2-16-gfc8670d1f72b
but the commit says fc8670d1f72b746ff3a5fe441f1fca4c4dba0e6f.

So the hashes are different, which led me to conclude that one referred to
the host kernel and the other to the kernel under test.

Where does the string "media_v5.1-2-16-gfc8670d1f72b" come from?

>>>
>>>
>>> 1  | qemu                   | arm64 | 118 total: 118 PASS   0 FAIL   0 SKIP
>>> 2  | qemu                   | arm   | 118 total: 118 PASS   0 FAIL   0 SKIP  
>>
>> Even if everything was OK, I think it would still be useful to have a link
>> to the full test report.
> 
> Yes, that is essentially the same issue as with the v4l-utils
> version as I described above.  The detailed results show a link
> to the console output, which isn't just a clean v4l2-compliance
> log but it's better than nothing.
> 
> We may also add a feature to publish some files alongside the
> parsed test results, and in the case of v4l2-compliance it would
> typically be the plain output of the test suite that developers
> are familiar with.  It's not a supported feature right now as
> only the raw console log is sent from the device to the database.
> 
> Best wishes,
> Guillaume
> 

Regards,

	Hans