In 51f89902 (slapos/cli/entry.py, slapos/cli/format.py, slapos/format.py:
Do not double the log output when "slapos node boot" failed., 2017-05-11)
we fixed a problem that every failure during boot causing twice more output
(which filled up the disk very fast), but a side effect was that using
--log-file no longer cause messages to also go to the console.

The "Always send higher-level messages to the console via stderr" comment
in entry.py seems that this was always intentional to output messages on
the console.

Keep the same "don't configure again logging system when it's already
configured", but implemented in a slightly different way, because since
a FileHandler is a StreamHandler, the check of already having a
StreamHandler was true also for the case where --log-file added a
FileHander (a few lines above, also in configure_logging method)

With nexedi/slapos!862 we no longer need this

See merge request nexedi/slapos.core!270

   Only assert the essential, otherwise any minor change on the zpt is a useless diff that pops up

  Mock NotificationTool_getDocumentValue to ensure the None as return value, in order to not find eventual valid notification messages with random content from other tests or dev environment.

  No need to assert entire message, it is enought to assert message isn't empty and assert a subset of it.

   This allows extend values w/o duplicate the whole test.

   Simplify the extensions on testSlapOSConfigurator to the expected bt5 and modules list

   This allow us to translate and re-run the tests with real notification messages.

   It ensure that selected payment is from the proper user.

  Also, keep paths specific to not conflict with extra business templates. We don't
  want a polluted list after the installation of extra bt5 on development or
  project specific environments.

   This allow introduce minor custom changes on project specific context w/o re-implement the whole test.

   This makes easy to overwrite and custom the list on sub-level projects

we no longer use python setup.py test

We were doing this because we use to run tests with `python setup.py test`
which automatically sets PYTHONPATH, so it was affecting sub processes
started by standalone.

Nowadays we no longer use this method of running tests, so we are not really
supposed to use PYTHONPATH, so we don't unset this environment variable,
there might be valid cases where users want to force a different python path
(but I can not think of any) and it simplifies the code a bit.

    Reported-by: Romain

Instead of letting slapos commands start supervisord in daemon mode, run it
as a program from standalone supervisord daemon.

The $INSTANCE/etc/supervisord.conf will be created by `slapos node instance`
the first time user runs it, until this, the service will restart in a loop,
complaining that config file is not found.

Also review stop to expose timeout argument and use a longer timeout by
default (and cleanup some unused imports).

I guess this code was using python -S to try to be independant from anything
that can happen in site.py, but that's not needed when running on testnode.

pygolang.pymain does not support -S and we are moving tests environments to
use pygolang.pymain, so let's remove this if we don't need it.

- also write pid file to prevent race condition
 - check user is root

See merge request nexedi/slapos.core!267

   Defined on the Constraint

   Include more verbose to the scripts to make debug easier.

See merge request nexedi/slapos.core!268

When comparing OS to see if they match for binary cache, we only compare
the lower case distribution name because sometimes "Debian' is
capitalized and sometimes not.

   It helps multi-account users to identify better where the notifications are coming from.