Previously this was '.sdimg'. But due to this not being a standard fileformat,
and unfamiliar to Debian users, the decision was made so standardize it on a
custom 'filename-mender.img' image naming convention.
Changelog: Title
Signed-off-by: Ole Petter <ole.orhagen@northern.tech>
The current execution of the acceptance tests is too long, in some cases
reaching the GitLab timeout of this project of 2 hours.
Rework the Pipeline to use a template and rework the script to allow
running tests for only one device type.
Changelog: None
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
The "high privileged" keys will only be available in protected branches,
so we won't be able to publish from a PR even if we run the manual job.
Changelog: None
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
The S3 path now follows the pattern:
/mender/mender-convert/uboot/<dev-family>/<dev-config>-<uboot-ver>.tar
For RPi and BBB these are automatically published form the CI Pipeline
of repo mender-convert-integration-scripts while for RockPro64 they have
been manually moved (as we expect this to be a one time integration).
Reworked a bit also the *_BINARIES variable so that we can keep the
u-boot version in the base config files and let the childs only define
the a config name.
Changelog: None
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
It turns out we will not use this after all, because the artifact
won't work properly without `mender setup`, which requires a live
device. We will revisit this and use a different approach once the
`mender snapshot` feature has been developed and merged.
This reverts commit 4cfc7bc648.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
This brings the rewrite of mender-convert into master, which will
become mender-convert 2.0.0.
This is essentially a merge with `-s theirs`. I also manually checked
there were no omitted changes from the existing master.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
This variable is actually used to detect that we are building with
Yocto. With the mender-convert rewrite, the relevant section started
working in mender-convert as well, so remove this, and also remove the
check in mender-image-tests.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
This makes it a bit more explicit that it is part of the testing
files. It follows the same pattern as in meta-mender with pulling
automatically, and using `--no-pull` to disable it.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
This avoid problems when running the tests locally. Technically it
will run the Raspberry Pi test twice on Gitlab, once with the prebuilt
image, and once with a test image, but the two configurations are not
identical, so I reckon that's alright.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
This also means using the new mender-gitlab-tmp-storage bucket to
transfer artifacts between stages, because the built in artifact
mechanism has a size cap that we are exceeding.
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
Make it clear that these are partition sizes, not filesystem sizes
(mainly important for the rootfs), and fix the ambiguous placement of
"x 2" (the existing placement made it look like there were 3892
partitions of 2 MiB each).
Changelog: None
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
* `IMAGE_ROOTFS_SIZE` - The base size of the rootfs. The other two
variables modify this value
* `IMAGE_ROOTFS_EXTRA_SIZE` - The amount of free space to add to the
base size of the rootfs
* `IMAGE_OVERHEAD_FACTOR` - Factor determining the amount of free
space to add to the base size of the rootfs
The final size will be the largest of the two calculations. Please see
the `mender_convert_config` file comments for more information.
Changelog: Commit
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
Not using this option produces a more realistic result, because block
padding has to be taken into account when calculating the totals. It
is not a perfect method however, since the calculation becomes
dependent on the block size of the filesystem on the build machine.
4096 is a *very* common size though, and almost certainly will be the
size used both on the build host and in the resulting filesystem.
Changelog: Changed the calculation of occupied space on the rootfs
partition to a more accurate formula.
Signed-off-by: Kristian Amlie <kristian.amlie@northern.tech>
Previously there was a mix between two, and four spaces.
This commits unifies all scripts to use two-spaces for indentation.
Changelog: None
Signed-off-by: Ole Petter <ole.orhagen@northern.tech>
Previously some scripts were calling env and bash, some were only calling bash.
This commit unifies the scripts to all use `#! /usr/bin/env bash`
Changelog: None
Signed-off-by: Ole Petter <ole.orhagen@northern.tech>
Ole Petter
Lluis Campos
Fabio Tranchitella
Kristian Amlie
Kristian Amlie
Alf-Rune Siqveland
Mirza Krak