[SSSD] [RFC] Add basic support for CI test execution

Lukas Slebodnik lslebodn at redhat.com
Wed Jun 18 12:17:56 UTC 2014


On (18/06/14 14:56), Nikolai Kondrashov wrote:
>On 06/18/2014 02:45 PM, Lukas Slebodnik wrote:
>>On (18/06/14 14:33), Nikolai Kondrashov wrote:
>>>On 06/18/2014 10:58 AM, Lukas Slebodnik wrote:
>>>>On (17/06/14 20:24), Nikolai Kondrashov wrote:
>>>>>On 06/11/2014 02:52 PM, Nikolai Kondrashov wrote:
>>>>>>On 06/11/2014 10:52 AM, Lukas Slebodnik wrote:
>>>>>>>>+        if [ "`xqilla -i \"\$f\" /dev/stdin \
>>>>>>>             Could you use xmllint instead of xqilla.
>>>>>>>We need xmllint for building manual pages. I didn't try it, but here is the 2nd
>>>>>>>link from quick google search.
>>>>>>>http://stackoverflow.com/questions/15461737/how-to-execute-xpath-one-liners-from-shell
>>>>>>
>>>>>>Ah, I didn't know xmllint could that. Thank you.
>>>>>>I tested it and will change to xmllint in the next version.
>>>>>
>>>>>Unfortunately, xmllint version in RHEL6 doesn't support that. So, although I
>>>>>already implemented its use, I'll have to go back to xqilla, which luckily
>>>>>exists on RHEL6.
>>>>"XPath tool" is necessary for extraction count of warnings from scan-build.
>>>>scan-build is not available in rhel6 it is only in epel6 and there is the same
>>>>version as in fedora [1]
>>>
>>>Well, we can't avoid using EPEL anyway, or can we?
>>>We need it for lcov and mock as well.
>>It is not problem with epel itself.
>
>Ah, good.
>
>>>>Could you explain why we need to run static analysis with the same version of
>>>>clang/llvm twice?
>>>
>>>We don't and we discussed this before. I just don't want to complicate the
>>>first release of CI and further postpone its use. Let's start with this. I'll
>>>work on separating distro-independent parts (including mock) afterwards.
>>>
>>If you prefer to have distro specific parts in another version of CI script
>>then static analysis should be also in different part (maybe 3rd) as well
>>and 1st version of CI scipt will be much simpler.
>
>I'll try to keep it in a single script, but provide a way to select
>all (the default), distro-dependent and distro-independent parts, so we can
>run them on separate builders. This will be orthogonal to selecting the
>magnitude of the test set.
>
>Did I understand you right, that you would like to run mock more often than
>the analyzer?
yes.
mock should be run on rhel/fedora.
and static analysis just once. (preferably on latest fedora). Static analysis
is quite slow it will reduce testing time for other distribution specific
builders.

>
>>>At least, as a minor positive side effect, this would allow executing most of
>>>the tests on RHEL6 too.
>>What kind of test do you mean? Is ther any problematic test on RHEL6?
>
>I mean dropping xqilla dependency would mean results of static analysis won't
>be verified on RHEL6 and it seems having that there would be slightly more
>beneficial than avoiding xqilla.
If you plan to "parse" other xml-log files on rhel6 it will be better to use
xqilla.

LS



More information about the sssd-devel mailing list