URL:
https://github.com/SSSD/sssd/pull/139
Title: #139: Initial revision of sssd pytest framework
lslebodn commented:
"""
On (03/02/17 00:39), Jakub Hrozek wrote:
On Thu, Feb 02, 2017 at 12:48:11PM -0800, lslebodn wrote:
> On (02/02/17 08:33), Jakub Hrozek wrote:
> >On Thu, Feb 02, 2017 at 08:29:15AM -0800, lslebodn wrote:
> >> On (02/02/17 08:14), Jakub Hrozek wrote:
> >> >On Tue, Jan 31, 2017 at 12:18:44AM -0800, fidencio wrote:
> >> >> Looking at the patches I see this can be a completely new library
instead of being part of SSSD.
> >> >> Knowing this it does make sense to have it as a submodule or even a
completely external library (but still under SSSD group).
> >> >>
> >> >> I'd like to know the opinion of the more experienced developers
about this.
> >> >
> >> >I also like the idea of a git submodule, because IIRC these tests are
> >> >supposed to be used kind of separately by downstream.
> >> >
> >> May I know why git submodule is better?
> >>
> >> The idea is that all test will pass with upstream
> >> so upstream distributions should use the version from tarball (+ patches)
> >>
> >> Or could you describe your use-case why git-submodule is better?
> >
> >I was under the impression that the tests should also be usable by
> >downstream. The submodule would make it possible to git clone just the
> >tests w/o the rest of sssd.
> >
> And how will you know which git hash from "submodule repo" should be used
> for testing specific version of sssd in downstream ?
I already keep the rhel- branches in my tree, I was thinking we might
expose them on some internal git server and collaborate there.
The intention of the effort is upstream first testing.
So the test need to pass in upstream first.
Could you explain how git submodule is related to "rhel- branches"?
I cannot see any benefit. You will need to have
the same branches n main repo and also in submodule repo.
This is the problem I am thinking about -- the tests might need
patching
when we add some patch to the 'downstream git branch'
Could you explain why
it would require patching?
It would be backported from upstream; the same as fix itself.
or we might just
add a test for a bug we fix in the downstream branch.
IMHO, Upstream first == fix will be backported with test.
But maybe it's just my naive idea.
On the other hand, the developers that mostly work on the tests (the
current QE team) might be adding test cases on their own.
But I see a problem here; upstream will fix patches also in stable
branches but an author of tests will not be aware of such think
They will not know where it should be backported.
So stable branches needn't have code coverage.
And I cannot see a problem why members of QE could not have
rights to push. We might ask how it is solved in 389ds.
Becasue IIRC QEs have right to push patches.
Summary:
maybe I am still missing something but I cannot see any benefit
of using git submodule.
LS
"""
See the full comment at
https://github.com/SSSD/sssd/pull/139#issuecomment-277936874