On (14/06/14 12:01), Nikolai Kondrashov wrote:
On 06/14/2014 11:47 AM, Lukas Slebodnik wrote:
>On (14/06/14 10:59), Nikolai Kondrashov wrote:
>>Hi Lukas,
>>
>>I've removed most of the discussion since we agreed to seek compromise on
>>dependencies and my answer below removes the need to discuss the rest.
>>
>>On 06/14/2014 01:21 AM, Lukas Slebodnik wrote:
>>>>I don't want to complicate the current script further by inventing
something
>>>>else above two sets of tests. If using Epoxy works out, it would allow
>>>>executing arbitrary tests and we'll be able to put mock there. For now
I think
>>>>building directly on the distro will be good enough. We don't have to
>>>>implement everything at once.
>>>I am not sure what is Epoxy. My google foo didn't help me. I do not see
>>>benefit in script which cannot be easily extensible.
>>>(or it can be but you have a reason to do not extend it).
>>>Your script can be used in temporary version of CI and probably is.
>>>If we want to have script in upstream I will prefer to have easily extensible
>>>version. If Epoxy is the right approach I will rather wait for next version.
>>
>>I've referred to Epoxy from the commit message - it's a Bash test
framework of
>>mine, which I'd like to use in CI in the future. It doesn't matter what it
is
>>really for this discussion.
>>
>>In the heat of the argument I somehow lost the track of what's important and
>>what's not, sorry. I was probably tired. I already have Clang running on
every
>>distro, taking 5 minutes and producing almost identical results everywhere, so
>>it is silly to argue against adding mock which will probably take half that
>>time.
>It is not so big overhead. It took 2 minutes on my machine. What I want to
>achieve is to test building of sssd in clean environment.
>I can see two options:
> 1) create clean VM for each test.
> 2) test with mock.
>
>Your idea was to use this script not only n CI machines, but also for
>developers on local machines. Testing in mock is more suitable for developers
>and has lower overhead.
Sure.
>You mentioned running clang on every distro. I do not see any benefit running
>all set of test on all distributions. Static analysis should not depend on
>platform specific details. The only diference can be in version of clang. Eac
>platform can have differet version of clang and newer clang has usualy better
>results.
I agree, but this will do for now.
>I also do not see a benefit of running all test after each commit. Tests with
>bigger overhead should be run per patchset.
> e.g. static analysis
I'll provide a way to choose what set of tests (full/essential) to run and
whether it should be run on all or just the last commit, upon submission.
You will
need to modify script anyway to address another comments in thread.
If you prefer to have this change in another patch I will be fine.
I am not sure about testing build of sssd in clean environment.
Will we create clean VM for each test or will we test with mock?
I can write patch for test in mock myself if you don't have experiences with it
LS