Re: [PATCH 2/3] services: Add wpa-supplicant-service.

["Tomáš Čech" <sleep_walker@gnu.org>]

I thought I sent reply already but I probably did not thanks to
connection problems ;-)

On Wed, 09 Nov 2016 23:32:52 +0100,
Ludovic Courtès wrote:
> 
> Tomáš Čech <sleep_walker@gnu.org> skribis:
> 
> > * gnu/services/networking.scm (wpa-supplicant-service): New procedure.
> > (wpa-supplicant-service-type): New variable.
> > (wpa-supplicant-shepherd-service): New procedure.
> > * doc/guix.texi (Networking Services): Document it.
> 
> Could you remove the ‘wpa-supplicant-service’ procedure and simply
> expose and document ‘wpa-supplicant-service-type’?  So users would
> write:
> 
>   (service wpa-supplicant-service-type wpa-supplicant)
> 
> instead of:
> 
>   (wpa-supplicant-service)
> 
> I’m trying to incrementally get us to do that for all services, because
> I think it’s a bit clearer (one immediately sees what type of
> configuration value is used.)

I can't say whether it is clearer or not from language or design
perspective, but after the change I had to have a look on your mail to
see how it is done properly so it is less user friendly. Nevertheless
done.

> 
> > +@cindex WPA Supplicant
> > +@deffn {Scheme Procedure} wpa-supplicant-service @
> > +       [#:wpa-supplicant @var{wpa-supplicant}]
> > +Return a service that runs @url{https://w1.fi/wpa_supplicant/,WPA
> > +supplicant}, an authentication daemon required to authenticate against
> > +encrypted WiFi or ethernet networks.
> 
> Could you also mention that wpa_supplicant is started to listen on a
> D-Bus interface?

Done.

> > +         (requirement '(user-processes dbus-system loopback))
> > +         (start #~(make-forkexec-constructor
> > +                   (list (string-append #$wpa-supplicant
> > +                                        "/sbin/wpa_supplicant")
> > +                         "-u" "-B")))
> 
> I think it would be more reliable to use
> “-P/var/run/wpa_supplicant.pid”, and thus
> 
>   #:pid-file "/var/run/wpa_supplicant.pid"
> 
> OK with changes along these lines, thanks!

Good idea!



> OK!

> So does Connman work as expected now?  I guess it was all broken before?

It works better with WiFi after this change but it is still far from perfect.

1] it still can be controlled only by root (it requires D-Bus policy
   to be installed properly and I need to figure out where and how)
2] connman-vpnd is not started - I'll probably make another service for that
3] there needs to be created some configuration mechanism for features
   like enabling/disabling dns proxy (which is IMHO killer feature of
   connman) or ignoring network devices
4] connman is harder to configure without other clients so I'm working
   on connman-json-client (curses) and connman-ui (GTK3)

Best regards,

S_W