GitLab

The original check was that master can be pushed only with develop
branch. It is what we mostly want but there are cases when we do not
update develop and only master that are valid as well. The example of
such operation is merge of develop to master, which is always
fast-forward due to our workflow.

This halts push only if master commit is not ancestor of develop instead
of halting it always when develop is not being updated.

This reverts commit 14abc044.

This reverts Foris part as well. The reForis part was reverted in
33ab8eff.
It turns out that my assumption that both Foris and reForis behave the
same that lead to only partial revert was invalid and Foris as it seems
it all right. This matches the behavior of reForis.

The second patch is fix. This backports patch that fixes issues with
scripts terminating but not correctly closing all outputs results into
hang.

The first patch is a feature. The turris-build repository contains
second part of it and it is already merged. It is not desirable to
release new version of updater because of other changes thus backporting
only this patch is a good solution.

Similarly to ebc12ecdb66a9261ba3d5cd06e87588764ac53aea2c64bd4c226f13cc2bf77d4

The `--` there was originally intended to disable processing of options
and thus protect against situation where password starts with dash. It
turns out that python processes arguments in such a way that anything
after command specification is automatically considered as argument and
not as option for python interpreter.

This also sets it to explicitly call Python3 compared to python as that
might call Python 2 that is not in package's dependencies and thus the
code might break if it would be called with it.

In recent update to version 2.2, I forgot to remove patch, which was
backported to fix issues which happened on blue Turris, powerpc.

Fixes: b508118a ("sentinel-minipot:
update to version 2.2")

- Update copyright
- Since version 2.2, there are required two dependencies: logc and
logc-argp.

The previous tarball is no longer available and it was necessary to push
a new container due to recent GitLab wipe and create new files again.

The content is still the same.

Added:
- Scripts that use storage for Lighttpd temporally files

This reverts commit 14abc044.

The issue with reforis starting was that both reforis and lighttpd were
creating FCGI socket. Thus lighttpd created socket and send request to
it while reForis replaced that socket when it initialized and thus never
get the initial request. The removal of socket configuration solves the
issue and the patch added here is that fix.

Fixes: CVE-2021-33880

* Use color-scheme detection for theme
* Sync translations with Turris OS
* Improve cards transitions
* Fix headline in README

Changed:
* Increase persistent logs size to 100 MB in logrotate config
* Refactor srv init script
* Restart rainbow service instead of hardcoded set to white in srv init script

Fixed
* Handle missing /srv directory in srv init script
* Prevent to manipulate programmable LEDs on routers without rainbow command in srv init script

Going to /netdata resulted in CSS and JS not being loaded which made
netdata unusable.

* faster rendering of snapshots table
* unified management of temporary files
* used credentials file for samba

Fixed:
* networks: fix reading multiple interfaces of wan (e.g. bridge)
* wan: ipv6_address in 6in4 may contain prefix as well

* Add & update translations
* Replace `attack detection` by `threat detection`
* NPM audit fix

Term 'data collection' is no longer desired. We should replace it
wherever possible with terms like:
 - Turris Security
 - Turris Sentinel
 - Threat detection

Call cleanup of the database faster to allow smaller databases.

It is being generated, but the contract enforcing part was missing
altogether.

Note: reading and writing the *.md5 file is now done within a couple lines,
instead of being sprinkled across a few scripts.

Also, change URL in PKG_SOURCE_URL as the repository was moved

* support for remotes on Samba
* allow to keep remote snapshots forever

In version 0.2.0, there was added feature to provide list of countries
[1]. There was also added to setup_requires (in setup.py)
dependency/package: l18n. OpenWrt has a problem
issue with it as can be seen below:

Error:
  error: invalid command 'bdist_wheel'
  ----------------------------------------
  ERROR: Failed building wheel for l18n
ERROR: Failed to build one or more wheels

It can be solved by adding host package build depends as described here
[2], which solved it for me so I could compile it.

[1] turris/foris-controller/turris-timezone@0a799a24
[2] https://github.com/openwrt/packages/blob/2edbc4c805d61d859f5e2af84293cdfb051f7f56/lang/python/README.md#installing-host-side-python-packages

Added:
* utils: add `ip_network_address` helper function

Changed:
* router_notifications: change default "from" and "host" of custom smtp server to empty s      tring

Fixed:
* dns: fix setting custom forwarder port for non-TLS servers
* lan: update openvpn server push route configuration on lan update
* lan: fix crash on dhcp host which has multiple MAC addresses (contributed by ktetzlaff)
* lan: fix crash on negative dhcpv6 lease time; set it to 0 for now
* wan: add missing fw rule for 6to4
* zeroconf: Foris-controller won't fully start when something is running on port 5353

Added:
* hook to reload openvpn server settings

Fixes:
* maintain-{lighttpd,network}-restart, maintain-reboot: fix reading of IP addresses

Changed:
* maintain-{lighttpd,network}-restart, maintain-reboot: Set mqtt as the only message bus used

Issue foris-controller#146 introduced new way of storing IP address and
netmask for lan, to be more aligned with the way LuCI stores it.

However it breaks several maintenance scripts, which in turn won't send
notification to re/foris to show the "reloading network..." spinner.

This fixes reading of IP address from uci.