URL:
https://github.com/SSSD/sssd/pull/511
Title: #511: Do not shutdown KCM/Secrets responders when activities are happening ...
fidencio commented:
"""
why do the first two patches fix code that the subsequent patches
refactor? Is it to make it clear where the issue is? Or did you consider backporting the
first two patches to some older branches?
It was to make clear that the issues has been fixed in the first two patches. The rest of
the series is "just" refactoring the code. IMO it's just easier for the
"future fidencio" to bisect the some issue and understand why things where done
in this way in the past.
So, answering your second question ... No, when I wrote the patches I didn't think of
backporting them to any specific branch ... but had in mind that distros could choose to
backport just those 2 patches instead of the whole refactor.
I see you added a new dependency python-psutils but only is_running()
is used. Have you considered implementing a simple version of is_running internally? I
think just checking for the same PID and the same executable and the same running time
would be quite OK.
python-psutils was there for free (on all supported OSes in our CI) and that's the
reason I've decided to use that.
IMO the effort of maintaining our on version of "is_running" (our anything else
already provided by a package that is there from el6 to latest fedora) is not worth it,
but if that's your preference ... I can write up my own and use it.
The tests you added are nice and it's really great that the
coverage is on par with the code, but the tests are also slow. 90 and 60 seconds
respectively is quite a lot. I wonder if you could send a mail to sssd-devel about this in
case anyone would be irritated by slow tests. I think it doesn't matter how long
Jenkins is chewing the tests, but if I was hacking on sssd-secrets, I wouldn't like to
wait for over two minutes for tests to complete the make intgcheck-run.
I'll send an email to the sssd-devel about this.
Another option would be to allow responders to run for less than 60 seconds (the minimum
default time) ... let's say change it to 10 and then those tests would be way
faster.
Would those changes be welcome? (I'll wait for your answer before dropping an email to
sssd-devel).
"""
See the full comment at
https://github.com/SSSD/sssd/pull/511#issuecomment-372477710