GitLab

* The get_remote_config_value returns tupple.
* The config callback should be called after the module is ready.

* It only takes a name instead of spec file name.
* The tests are just recycled tests for the previous, with some code
  reuse tricks (so the actual test code is shared between them).

When closing a ModuleCCSession, a 'i am stopping' message is now sent to the ConfigManager, which in turn informs Cmdctl (which then causes bindctl to update its list).

This causes stopped modules to disappear from bindctl until such time that they are enabled or started again, so that config updates and most importantly, commands, do not cause weird timeouts (bindctl will immediately inform you of modules not running)

The thing should be converted to string properly

this bug resulted in the symptom that when removing from the default value, the removal wasn't stored

this problem surfaced when 1246 got merged; the ModuleCCSession start in bind10 actually relied on some environment variable leaking out (which 1246 fixed).
This change makes it an explicit optional argument to ModuleCCSession (which can still be overridden by the environment variable)

Boss now waits to receive a message from the configuration manager
to say that it is running before starting other processes.

I had not noticed the -p option, which doesn't really make sense and should perhaps be removed, but that one actually does work with relative paths, so the original 'fix' was bad

plus a couple of fixes to make relative paths work in bind10 -c

instead of dynamically creating them so that all of in-src separate-builddir
and distcheck will succeed.