[Qa-jenkins-dev] testing initsystems Re: Why sysvinit?

[Holger Levsen]

Hi Simon,

On Sonntag, 6. März 2016, Simon Richter wrote:
> > I'd be very glad to support you with setting up these tests on
> > jenkins.debian.net, where we already test similar things and which has
> > ressources to do a lot more tests.
> Cool, thanks!

to answer some questions you asked me in private:

a.) you don't need to be subscribed to this or any other mailinglist to use 
and develop code used on jenkins.d.n, though *sending* mails to this list 
(instead of me or other maintainers) is appreciated.

b.) yes, you can write and maintain your own code there, just make sure it 
uses sensible namespaces. then clone jenkins.debian.net.git, write patches to 
it and notify me of the commits I should cherry-pick. As long as you only 
touch your stuff, eg bin/initsystem-tests.py or job-cfg/initsystem-tests.yaml, 
I will happily+"blindly" merge and deploy it. Ususally just highlighting me on 
irc with the commit hash(s) in question is enough.

Technically also Helmut and Mattia can merge code but in practice currently 
I'm the one doing all the reviews and merging. (I do more/deeper reviews on 
some stuff but usually I go by "merge, revert is easy" :)

> > What virtualisation do you think you'll need for your tests? QEMU/kvms?
> > or will chroots be enough?
> 
> I think both VMs and chroots will be needed.

ok

I have never tried it (because it's rather new), but vmdebootstrap (available 
also in jessie-backports) seems to be a useful tool here.

> Most of the tests will run fine in a chroot (especially debootstrap
> tests), but it'd feel incomplete if we didn't at least try to actually
> boot the system and see whether it comes up.
> 
> Interesting challenges:
> 
>  - Can we boot a VM from file system images that are artifacts from
> another build?

yes, but I'm not sure it' worth it.

>  - How to verify that the system in the VM booted correctly

ssh in?

> Current idea: build a custom service that runs pstree, magically sends
> it to the build log, and then compare the output against the expected
> value.

keep it simple.

>  - Can we somehow generate an APT repository from an Alioth project?

sure. check /home/groups/reproducible/htdocs/debian/Makefile on git.d.o

> When someone commits to the sysvinit repo, I'd like to build the
> packages (probably easy),

yup

> then create something that debootstrap can
> access, ideally as an artifact (tar archive of repo?)

upload the packages to a custom/on demand/temp repo?

> I think step one should be simple transition tests:
> 
> 1. install a system with debootstrap
> 2. transition to sysvinit
> 3. transition back to systemd
> 
> These should probably be three different projects, so we can reuse the
> installed system at any point as an artifact, but that's something
> someone more experienced with Jenkins might know better.

I'd implement these three different projects first. And then start thinking 
about those next steps.


cheers,
	Holger
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 828 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.alioth.debian.org/pipermail/qa-jenkins-dev/attachments/20160307/90039bcd/attachment.sig>