[gpsd-dev] Configuring GitLab CI (was Upcoming migration to GitLab)

[Jason Azze]

On Sat, May 18, 2019, at 4:49 PM, Jason Azze wrote:
> On Sat, May 18, 2019, at 10:05 AM, Eric S. Raymond wrote:
> > Christian Gagneraud <address@hidden>:
> > > Do you have any plan for the CI jobs? (build, test, regression, sanity
> > > check, ...)
> > > The best, IMHO, is to run your own CI runner if you have the resources.
> > 
> > We'll migrate them over of course - such things are not difficult to do on
> > GitLab.  But I didn't write those bits and don't know the details of what
> > they're doing.
> > 
> > Would the responsible party speak up, please?
> 
> I run CI jobs on a private Jenkins instance, but I'd be interested in 
> duplicating the functionality in the GitLab CI environment in the same 
> way we do for the NTPsec project.

This weekend I made a fork of the project in order to test integration with 
GitLab's CI/CD infrastructure.

https://gitlab.com/jason-azze/gpsd/blob/master/.gitlab-ci.yml

 I extended the .gitlab-ci.yml so it would run a build on each push. This 
worked nicely.

Then I added a `scons check` step to run tests. This didn't work. The 
regression tests probably need something a Docker container can't (or won't) 
provide, like access to the D-BUS system bus as root. Because, as the code 
comment says in gpsd.c

    /* we need to connect to dbus as root */

So, the test run attempts look like...

Testing the daemon...
Processing test/daemon/GPSmap-76S.log
gpsd:ERROR: unable to connect to the DBUS system bus

That's as far as I got before I went outside to play.

I will try next on a runner that I control (rather than on GitLab shared 
runners), unless someone beats me to it.

-- 
Jason Azze