Allow users to migrate custom files between installs.

Users can create files like this:

  $ cat /etc/rauc/migrate.d/git_config.conf
  /etc/gitconfig
  /home/root/.gitconfig

Which will result in the listed files being migrated to the new
RAUC slot.

Signed-off-by: Leonard Göhrs <[email protected]>

Signed-off-by: Leonard Göhrs <[email protected]>

Fix an optional shellcheck warning to use double square brackets for bash
scripts:

  In meta-lxatac-software/recipes-core/bundles/files/hook.sh line 26:
  			if [ "${bundle_hash}" == "${cert_hash}" ]; then
                             ^-- SC2292 (style): Prefer [[ ]] over [ ] for tests in Bash/Ksh.

  In meta-lxatac-software/recipes-core/bundles/files/hook.sh line 39:
  	if [ ! -f "$1" ]; then
             ^-----------^ SC2292 (style): Prefer [[ ]] over [ ] for tests in Bash/Ksh.

Signed-off-by: Rouven Czerwinski <[email protected]>

Fix an optional shellcheck warning:

  In meta-lxatac-software/recipes-core/bundles/files/hook.sh line 60:
  			migrate "$x"
                                   ^-- SC2250 (style): Prefer putting braces around variable references even when not strictly required.

Signed-off-by: Rouven Czerwinski <[email protected]>

…ellcheck

Fix optional shellcheck warnings for hook.sh

meta-lxatac-software: bundles: hook.sh: migrate user-specified files

The content of the /srv/www directory is exported by the tacd via HTTP.
If this directory does not exist the user gets a (true, but somewhat
confusing) 404 Not Found error if they try to access /srv via HTTP.

Usually a tmpfiles.d entry would be the way to go to create this directory,
but it causes issues for directories on the /srv file system.
See this comment copied from a similar commit for atftp.service:

The tmpfiles.d service runs very early in the boot process and even before
system-update.target. This means it would clutter the /srv directory in the
rootfs if the /srv partition was not yet mounted - because it was not generated
yet by systemd-repart.

Signed-off-by: Leonard Göhrs <[email protected]>

meta-lxatac-software: tacd: create /srv/www in tacd.service

This should make sure that we do not forget to update the DISTRO_VERSION
and DISTRO_CODENAME when a new tag is created or a new yocto version branch
is started.

Signed-off-by: Leonard Göhrs <[email protected]>

CI: add distribution version consistency checks

The gpiofind command was removed, so the tacd-failsafe command would
currently fail (which is kind of ironic).

Instead gpioset can be directly supplied with the GPIO names.
The GPIOs are not guaranteed to keep the assigned value once gpioset
exits, so we just keep it running for some time before trying to restart
the tacd.

Signed-off-by: Leonard Göhrs <[email protected]>

meta-lxatac-software: tacd: failsafe: gpiofind no longer exists

To follow a well honored tradition our first CI job we added failed
instantly when it landed in the main branch.
Now that the traditions have been honored we can fix it.

There were multiple things wrong with the previous version regarding
CI jobs running for a branch or tag (instead of jobs running for a pull
request):

  - The GITHUB_REF variable contains a name like "refs/heads/mickledore",
    not just "mickledore". We want GITHUB_REF_NAME.
  - The GITHUB_BASE_REF variable is also set for non pull request jobs,
    but it is set to an empy string. Check for that to decide if we are
    running in "pull request mode".
  - When running for a tag the branch name check for the os codename
    does not make sense. Skip it in that case.

Fixes: 7fa11b2 ("CI: add distribution version consistency checks")
Signed-off-by: Leonard Göhrs <[email protected]>

CI: fix distribution version consistency checks

meta-oe provides android-tools 29.0.6 in nanbield, which should be more
than new enough to suite our needs.

Signed-off-by: Leonard Göhrs <[email protected]>

In addition to latest-and-greatest being the way to go this also fixes
issues with newer compilers complaining about uint32_t being used without
including stdint.h. See [1] for the respective upstream commit.

[1]: nxp-imx/mfgtools@24fd043

Signed-off-by: Leonard Göhrs <[email protected]>

The libiio version in meta-oe was updated and we no longer have to carry the
backported patch.

Likewise with the dt-utils patch and meta-rauc.

Our ".rootfs.ext4" -> ".ext4" workarounds in lxatac-core-bundle-base.bb and
emmc-image/lxatac/genimage.config are also no longer required.

The perf package now handles the libtraceevent dependency correctly and we
do not need that bbappend anymore.

Signed-off-by: Leonard Göhrs <[email protected]>

Update to Yocto 4.3 - Nanbield

The remaining fields are filled in when the user is actually created,
having it in these files may be missleading when changes to the file
do not propagate into the generated image.

Signed-off-by: Leonard Göhrs <[email protected]>

The tacos base install does not install / configure systemd-journal-upload,
but the Pengutronix-Internal derived images do.
Register the uid/gid here, as systemd-journal-upload is generally useful
when managing a fleet of TACs.

Signed-off-by: Leonard Göhrs <[email protected]>

meta-lxatac-software: passwd: reserve uid/gid for systemd-journal-upload

Signed-off-by: Leonard Göhrs <[email protected]>

…rsion

Currently the rauc version in rauc-1.11.1.bb is newer than the version
in rauc_git.bb. This is obviously an uncommon situation, but it made
us aware that we set RAUC_USE_DEVEL_VERSION=1 without a clear reason
or need.

Released RAUC has all the features we need. Use it.

Signed-off-by: Leonard Göhrs <[email protected]>

Currently the update notifications on the LXA TAC LCD and web interface
do not work because the tacd can not get informations about the RAUC
bundle on the update server via RAUC's DBus API.

This is because the process that RAUC spawns to handle the nbd bridge
for http streaming crashes due to an unitialized pointer access.

This patch is taken from RAUC pull request 1332 [1], which fixes the issue.

[1]: rauc/rauc#1332

Signed-off-by: Leonard Göhrs <[email protected]>

meta-lxatac-bsp: rauc: add patch to fix RAUC info / update notifications

This effectively bumps the Debian version to the most recent release at
that time.
This way we will not need to update the script with every Debian
release.

This also aligns the behavior with `container-start.sh` that simply
pulls `debian` if no container is present.
And this also defaults to `latest`.

Signed-off-by: Chris Fiege <[email protected]>

poky/meta-*: update to most recent nanbield branches

meta-lxatac-software/containers: Update to debian:latest

The official GitHub action runner client is written in C# and uses
(as far as I could figure out) the non-free(?) part of the .NET
Framework, making it hard to build.

This alternative client is written in Go and relatively easy to build.
We'll have to see how compatible it is in practice.

The need for nodejs on the TAC is due to all "uses:" steps for
runners being implemented in javascript (*sigh*).

To configure a runner you have to into the Actions setting of your GitHub
repository, and add a new self-hosted runner.
After configuring the OS and Architecture (Linux / ARM) you can use the
presented token to register the runner:

  $ mkdir -p /etc/github-act-runner
  $ cd /etc/github-act-runner
  $ github-act-runner configure \
      --name $(uname -n)
      --token 01234567899876543210012345678
      --labels dut-raspberrypi
      --url https://github.com/user/repository

Signed-off-by: Leonard Göhrs <[email protected]>

To configure a runner you have to into the CI/CD setting of your GitLab
repository, and add a new project runner.
After configuring a few aspects of the runner you are presented with a
token to use on the TAC:

  root@lxatac:~ gitlab-runner register
      --url https://gitlab.com
      --token glrt-01234567899876543210

Signed-off-by: Leonard Göhrs <[email protected]>

GitHub/GitLab Action/Pipeline Runners

Signed-off-by: Leonard Göhrs <[email protected]>

This was suggested by shellcheck.

Signed-off-by: Leonard Göhrs <[email protected]>

Barebox shell rules are not 1:1 equivalent to "/bin/sh" rules,
ignore this error.

Signed-off-by: Leonard Göhrs <[email protected]>

Signed-off-by: Leonard Göhrs <[email protected]>

…nings

Signed-off-by: Leonard Göhrs <[email protected]>

Signed-off-by: Leonard Göhrs <[email protected]>

Signed-off-by: Leonard Göhrs <[email protected]>

Signed-off-by: Leonard Göhrs <[email protected]>

…cripts

Signed-off-by: Leonard Göhrs <[email protected]>

This change is less mechanic than the previous tac-gadget shellcheck
change and required some more thinking, which is why it is split
out from the rest.

Signed-off-by: Leonard Göhrs <[email protected]>