Hi Mathieu,

Thanks a lot for taking so much time to help me out and review this
patch.

Following your previous mail, I'm currently writing a test. I may follow
NFS's tests rather than CUPS -- (some of) NFS' tests just checks from
within the marionette that the service has been started, while CUPS
verifies from the outside that an external service is correctly running.

While I do prefer the CUPS test -- verify at the "user level" that the
service is provided -- it seems very complex to me to test cachefilesd
at a high level with a binary decision and no false positive. Venues for
this could be:

- maybe an NFS mount with the fsc option when cachefilesd is not
  activated will fail -- I need to check that;

- or maybe mounting some NFS share with fsc, accessing it, and checking
  afterwards that the cache has been filled by cached data -- this is
  getting complex though, and probably not a clear-cut scenario.

I don't know how high you will set the bar and if a simple 'check that
the daemon is running' would be sufficient to you :)

Thanks a lot for the various documentation pointers and style
recommendation. I really appreciate all of this!

I will provide a comprehensive patch with your changes integrated, and a
test, as soon as i am happy with it -- or just give up writing the
perfect test.

Kind regards,
Jean-Baptiste