When user ran out of inotify watch, starting wrappers using
wait-for-files is not possible:

    AssertionError: Traceback (most recent call last):
      File "/srv/slapgrid/slappart15/tmp/tmpzurwmg4q/wrapper", line 13, in <module>
        sys.exit(slapos.recipe.librecipe.execute.generic_exec(['/bin/echo', 'done'], wait_list=['/srv/slapgrid/slappart15/tmp/tmpzurwmg4q/wait']))
      File "/srv/slapgrid/slappart15/srv/project/slapos-clean/slapos/recipe/librecipe/execute.py", line 100, in generic_exec
        _wait_files_creation(wait_list)
      File "/srv/slapgrid/slappart15/srv/project/slapos-clean/slapos/recipe/librecipe/execute.py", line 30, in _wait_files_creation
        watchdescriptors = {inotify.add_watch(dirname,
      File "/srv/slapgrid/slappart15/srv/project/slapos-clean/slapos/recipe/librecipe/execute.py", line 30, in <dictcomp>
        watchdescriptors = {inotify.add_watch(dirname,
      File "/srv/slapgrid/slappart15/srv/project/venv/lib/python3.9/site-packages/inotify_simple.py", line 110, in add_watch
        return _libc_call(_libc.inotify_add_watch, self.fileno(), fsencode(path), mask)
      File "/srv/slapgrid/slappart15/srv/project/venv/lib/python3.9/site-packages/inotify_simple.py", line 39, in _libc_call
        raise OSError(errno, os.strerror(errno))
    OSError: [Errno 28] No space left on device

This catches inotify errors and fallback to simple polling in that case.

Running logrotate -d on cluster with a lot of slaves results with execution
taking a lot of time (>10s, one some nodes ~20s), which leads to random
promise timeouts.

instance-slave-list.cfg.in is run by instance-frontend.cfg.in, which
implements that exact promise already.

Default software_type requests all required instances for the swarm.

See merge request nexedi/slapos!1264

Add getSpeed and getClimbRate to drone's API

Update FPS value to respect minimum sampling value of open62541 monitored items

Make sure to stop Pub/Sub server before stopping MavSDK thread

Now the default software-type request a list of drone software-type

keep old versions for python2

also sort versions and remove comments that make it hard to sort
versions and are not so helpful:
 - we can find the dependents of a package by looking in
  *eggs/*/EGG-INFO/requires.txt
 - nowadays, many python packages no longer have setup.py and wheels
  is becoming standard.

jsonschema egg now needs typingextension egg so we needed to pin it

In order to cleanup, I moved all eggs at once.

This reverts commit b3afa3fe.

This change unexpectedly impacted ERP5 deployment.

See merge request nexedi/slapos!1327

Let's pick up all fixes and improvements accumulated upstream to be used together with Zope4/Python2.

/reviewed-by @jerome, @levin.zimmermann
/reviewed-on nexedi/slapos!1332

See merge request nexedi/slapos!1331

fonts ipaex-fonts wouldn't download on version 1.0.210, and aren't
available in shacache for some reason.

ZEO 5.4.0 fixes several concurrency bugs and provides other
improvements. See changelog in https://pypi.org/project/ZEO/5.4.0/
for details.

Going from ZODB 5.6.0 to 5.8.0:

- removes leftovers of pre-MVCC read conflicts,
- relaxes assertion in tests that relate to NEO,
- brings up interface requirements in line with Wendelin.core 2 and WCFS,
- optimizes IO in fsrefs,
- fixes fsdump to match fsstats,
- provides new tests that exercise potential concurrency issues in
  storage implementations more thoroughly,
- provides other fixes and improvements.

Please see https://github.com/zopefoundation/ZODB/compare/5.6.0...5.8.0 and
https://pypi.org/project/ZODB/5.8.0/ for details.

Database password is now randomly generated.
Now test that the pre-filled password matches the mariadb url.

By inspecting current configuration stored in .data-disk-amount of the
partition configuration directory extend the conflict protection to really
observed configurations.

Adapt the README to follow real messages and explain a bit more the situation.

Patch nexedi/slapos@15e26abd explicitly declared
go version 1.17 for the 'helloworld' SR.

This was done at a time when we already had [golang1.17] in component/golang,
but the default was still go1.16. The explicit declaration allowed the 'helloworld' SR
to already switch to the newest go version.

Since patch nexedi/slapos@d32524c3
we use go1.18 as our default version. In order to allow 'helloworld' to use our newest
SlapOS supported go version, it's therefore sufficient nowadays to simply remove the
explicit declaration.

/reviewed-on nexedi/slapos!1330
/reviewed-by @kirr

Database password in stack/lamp is now randomly generated.

The ssh login directory stays at the outer HOME directory.
For more info, see the README.md

See merge request nexedi/slapos!1329

Reuse stack and drop overcomplicated rsyslogd configuration.