So that GitLab CI script (or any user, really) can customize the name of
the resulting Docker image.
Changelog: None
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
The goals of the re-write was to achieve the following:
1. It should be easy to extend the tool to support other boards or distributions
2. We should not compile code in the tool (rely on binaries built elsewhere)
compiling code increases complexity, due to requirement of toolchains etc.
3. The tool shall not be designed around specific hardware/platform types
- This is the case today with the usage of --device-type flag
4 The tool should be to convert images without knowing anything about the hardware/platform
relates to above 3.
5. Configuration interface should be simplified
- command line flags -> configuration files
6. Platform specific code shall be provided trough “hooks”, and are not part of the “core” mender-convert code
7. It shall be easy to extend functionality
- support for rootfs overlay to inject user applications/configurations
- ability to override how the Mender Artifact is generated (to be able to sign and include state-scripts)
8. Code structure should be modular
- Eases Maintenance and possibility of making isolated changes
Changelog: Title
Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
So there are other arguments that need to be passed to docker build,
namely toolchain related, therefore it is better to use the default
script and just override the image name.
Also, removing the Mender version from the GitLab pipeline (and hence
using the one defined in build-docker) removes one point of
inconsistency.
Changelog: None
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
So far the assumption has been that we are targeting an arm (32-bit),
when we are compiling the Mender client during docker container build.
To be able to support other platfors we need to parameterize this, and
this commit adds support for arm64.
Changelog: Title
Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
When installing the client, call "make install" from source directory
and then modify the default setup with the user input (tenant token,
server url, demo...), eliminating this way the duplication on the
install recipe.
Also, setting the default build version to 2.0.0 beta
Other minor changes along the way:
* Include liblzma build dependency
* Change --mender-client parameter to be optional
* Abort on error in convert-stage-4.sh
* Bugfix in getting server_cert from command line option
* Error reporting when using unsupported device
* Clean-up obsolete --help example
Changelog: Title
Signed-off-by: Lluis Campos <lluis.campos@northern.tech>
consistent and align with entry-point way of running mender-convert.
Changelog: None
Signed-off-by: Eystein Måløy Stenberg <eystein.maloy.stenberg@northern.tech>
It builds for both RPi3 and BBB, but there is a mount
warning for BBB, which may lead to it not working properly.
Changelog: Docker environment for running mender-convert
Signed-off-by: Eystein Måløy Stenberg <eystein.maloy.stenberg@northern.tech>