When using bootflow, the fallback device tree is being constructed from
the soc
(k3
) and board
(am62x
) environment variables:
Retrieving file: /dtbs/6.8.1+/k3-am62x.dtb
This file doesn't exist.
The environment variables default_device_tree
and
default_device_tree_arch
are set in the config and will result in
fdtfile
being set to ti/k3-am625-sk
when the findfdt
script is
run, which isn't the case at the moment.
Calling this script results in the correct device tree being retrieved:
Retrieving file: /dtbs/6.8.1+/ti/k3-am625-sk.dtb
Many boards are calling this script as part of their boot command. The am62x currently isn't. Rectify this so that booting works correctly.
Signed-off-by: Martyn Welch martyn.welch@collabora.com
Right, my question was more around how we are going to ensure this gets forwarded & fixed upstream?
Does this need to be forwarded upstream and a Forwarded:
tag ?
In !499 I tried to create a common base ospack to reduce code duplication. This MR has been started with the goal of to reducing code duplication.
There is a lot of duplicated code between the different ospacks; and there is no real place to start if one would like to start from a base ospack (the fixedfunction ospacks contain nginx so arguably they are not minimal enough to be a base image).
So let's create a base ospack and move the common ospack bits into a seperate recipe on the path to abstracting out the common parts of the image recipes into a common base.
Leaving as a draft as currently more refactoring is required.
I pushed a fix upstream for this issue: https://github.com/go-debos/debos/pull/479
the Python yaml library is very inefficient both CPU and memory wise
Can you quantify before vs after in the commit message?
Christopher Obbard (12577f2b) at 18 Oct 19:53
Christopher Obbard (12577f2b) at 18 Oct 16:34
Release linux version 6.3.7-1+apertis7
Source packages imported from Debian do not contain debian/build/version-info. This causes the update-metadata script to fail. Make the file optional so that the script continues to execute in these cases.
Signed-off-by: Christopher Obbard chris.obbard@collabora.com
Christopher Obbard (12d61d46) at 11 Oct 13:36
update-metadata: Don't bail if d/b/version-info isn't present
Currently the proposed-updates branches (i.e. those branches which are automatically being updated from upstream) are only built as source packages and no test binary is built in the merge request context. This means that we don't check if an automatically imported package update can be built; the only binary build is completed after the merge! Let's enable test builds of imported packages pre-merge to not let any build failures on import slip through the cracks and require manual work later.
Signed-off-by: Christopher Obbard chris.obbard@collabora.com
Allow cross-building packages by accepting a Debian architecture
from an environment variable named $ARCH
.
Allow build profiles to be passed to debuild by using the
environment variable $DEB_BUILD_PROFILES
.
First build the docker container (after cloning this branch):
docker build package-source-builder --build-arg RELEASE=v2023dev2 --tag apertis-v2023dev2-crossbuild
Cross-building the Kernel package for arm64 (after cloning the Kernel package tree):
git clone https://gitlab.apertis.org/pkg/linux.git
cd linux
pristine-lfs checkout --auto
docker run --rm -it \
-v ${PWD}:/mnt/pkg \
-w /mnt/pkg \
-e DEB_BUILD_OPTIONS=parallel=$(nproc) \
-e DEB_BUILD_PROFILES="pkg.linux.notools nodoc noudeb" \
-e ARCH=arm64 \
apertis-v2023dev2-crossbuild \
ci-buildpackage
Signed-off-by: Christopher Obbard chris.obbard@collabora.com
Let's close this merge request
Nothing has happened as yet. It'd be great if you could reproduce this in the latest everything ;-)
After the last update of fakemachine some regressions were introduced. In order to fix them update to the latest version, which only introduces low impact changes and have been tested in v2024dev1 for while now.
Christopher Obbard (94ed998b) at 23 Mar 22:18
Release golang-github-go-debos-fakemachine version 0.0.4-1+apertis0
... and 3 more commits
After the last update of fakemachine some regressions were introduced. In order to fix them update to the latest version, which only introduces low impact changes and have been tested in v2024dev1 for while now.
@wlozano We have had no changes on apertis/v2024
for the last few weeks. Can you take care of the backport now which should fix this issue ?