This moves the U-Boot environment migration from Omnia to generic
nor-update. We want to check if migrate is required every time we flash
the new version of U-Boot.

The result without argument returns $? and thus 1 that triggers error
and termination of nor-update. The zero has to be returned explicitly
there.

The ss command on Turris OS 5.x does not support the --cgroup argument.
It is available only since Turris OS 6.0.0.

This should enable foris-controller zeroconf
machinery working even when umdns is used
(it occupies port 5353) and makes it impossible
for foris-controller to generete zconf announcements

After this patch the umdns will foris-controller compatible
announcements.

This should add support for new zconf scheme.
Old scheme is still supported, so this should
be compatible even with older versions
of foris-controller.

Remove backported patch, which is included this release

Fixes: CVE-2021-46141, CVE-2021-46142

This file, which I removed is no longer used as it was superseded by
the package mox-dtb. It does not require to bump package.

Fixes: 26fba930 ("nor-update: Generalize
to work for all devices")

Foris Controller expects specific types in json output and doesn't use
the output that could produce invalid json, so let's temporally revert
one fix till Foris Controller is updated.

Drop all backported patches.

Upstream release notes for 2.8.0

* allow to specify samba protocol version in remote URL
* support export filter and overlay
* various small fixes (mainly to remote synchronization)

Tool mkfs.(v)fat is provided by package dosfstools and in OpenWrt commit [1], it was merged to single package. Because of that mkfs.vfat is no longer provided.
Let's use mkfs.fat.

[1] https://github.com/openwrt/packages/commit/5d032bce1f0eb259ba94e7d8bed1444a1236f8b0#diff-4ae5696ed12100b6b90f3e12991fcd054b35f4fbb3af426d5bacdda6f2261948

Added
* Tests

Changed
* Package structure refactored
* Update description

There is a conflict between uci functions provided by OpenWrt and
functions used by the script, so moving loading options to the end had
some unforeseen consequences.

As merging is currently blocked, you can read up detailed changes in
patchset attached. Once merged, I'll tag 2.8.0 with full description and
release 2.8.1 with potential fixes.

Michal Hrusecky authored 3 years ago and Karel Koci committed 3 years ago

Should make no real difference, just code tiny bit looks nicer.

Michal Hrusecky authored 3 years ago and Karel Koci committed 3 years ago

Although the idea is great, it is currently not supported in reForis and
would cause more troubles, so let's drop it for now.

Michal Hrusecky authored 3 years ago and Karel Koci committed 3 years ago

SFP interface is now called SFP, no longer eth1. Switch to DHCP client
only if we have no WAN port - only one module device. If we have a switch,
set WAN as bridge between SFP and eth0.

Fixes issue with GitHub Actions (CI):
Dirty patches detected, please refresh and review the diff

The LAN ports are on Turris 1.x in reverse order compared to switch
physical ports. We have to map them in reverse.

The unterminated VLANs were not previously migrated as they are not
terminated in CPU. This adds support for them as one user reported that
it is a setup he is using.

The `vid` is in default set by LuCI but if user configures stuff
directly it can use only `vlan` and in default that is used as `vid`.
This is probably a common way most of the advanced users configured
switch. We need to read vlan and use that as default to vid.

We missed this and of course we read the value so we need to have pin as
input not as output.

Fixes:
CVE-2021-2372
CVE-2021-2389

It seems that mirror returns HTML 404 and thus makes it not available to
download.

Some of our Turris packages does not provide a way to check version via
parameters --version, -version, etc.

Also, for now, we can not use packages which depends on meson as the
path for meson was changed, which was done as patch for turris-build
repository. Unfortunately, for us, we are using OpenWrt SDK and our
packages if it builds.

There is some issue while building this package and it could be
reproduced locally. It was sufficient to drop the row responsible for
l18n.

Let's see if our CI can build it as well.

Output:
Collecting l18n
  Using cached l18n-2021.3.tar.gz (50 kB)
    ERROR: Command errored out with exit status 1:
     command: /hbk/staging_dir/hostpkg/bin/python3.7 -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/hbk/tmp/pip-install-uup06pus/l18n/setup.py'"'"'; __file__='"'"'/hbk/tmp/pip-install-uup06pus/l18n/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(__file__);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' egg_info --egg-base /hbk/tmp/pip-pip-egg-info-jvy2c1hf
         cwd: /hbk/tmp/pip-install-uup06pus/l18n/
    Complete output (5 lines):
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/hbk/tmp/pip-install-uup06pus/l18n/setup.py", line 88, in <module>
        cmd_module = getattr(__import__(module, fromlist=[cmd]), cmd)
    ModuleNotFoundError: No module named 'wheel'
    ----------------------------------------
ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.
make[2]: *** [Makefile:46: /hbk/build_dir/target-aarch64_cortex-a53_musl/turris-timezone-v0.2.1/.built] Error 1
make[2]: Leaving directory '/hbk/feeds/turrispackages/utils/turris-timezone'

The commit 43c14e4c updated defintions
but not pacakge version.

This fixes issue with l18n as there is missing timezone - Pacific/Kanton