To make sure we have reliable and repeatable builds, the Dockerfile
now uses "ubuntu:noble" instead of "ubuntu:latest" to have a fixed,
known recipe for containerized builds.
Signed-off-by: Ed Beroset <beroset@ieee.org>
This updates to the latest ubuntu image. It also removes
the deprecated "python3-distutils" and adds "python3-packaging" and
"file" to the list of installed dependencies. This fixes#299 in which
libmce-glib would fail to compile with a Python error
"ModuleNotFoundError: No module named 'packaging'"
Signed-off-by: Ed Beroset <beroset@ieee.org>
46339cad31
introduced behaviour such that the 'current' state of `bblayers.conf` is copied to
a temporary file layers are updated and this bblayers.conf is overwritten with
the temporary file generated by `mktemp`.
However, this introduces an issue where an empty `bblayers.conf` is created if
the user decides to update layers perior to building (i.e. do
`./prepare-build.sh update` first) as the temporary file (created with `mktemp`)
is copied without any contents because `bblayers.conf` doesn't exist yet.
This empty file doesn't contain the `layer_line` pattern, so nothing will be
appended.
To fix this issue, do not generate an empty `bblayers.conf` file upon update if
it doesn't exist.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
There was a silly omission that caused the bblayers.conf file to be empty under
some conditions. This fixes that and restores full functionality.
Signed-off-by: Ed Beroset <beroset@ieee.org>
The previous patch was not quite correct. It only copied into the
tmpfile once but should have copied once per iteration of the for loop.
This fixes that error.
Signed-off-by: Ed Beroset <beroset@ieee.org>
This removes the requirement for the inplace script or any gawk features
by explicitly creating a tmpfile, operating on that, and then removing
the tmpfile when complete.
Signed-off-by: Ed Beroset <beroset@ieee.org>
This updates meta-open-embedded beyond the labeled mickledore version to
pick up the patch that corrects the current SRC_URI for polkit as well
as a few other fixes and improvements.
The upstream polkit change needed is this one:
0804e61bd7
Signed-off-by: Ed Beroset <beroset@ieee.org>
This commit merges 'apt' commands and cleanup into one Docker layer. This is crucial because cleanup within the same layer it's created is effective, unlike in separate layers. This change improves Docker image efficiency by reducing the overall size.
There have been increasing reports of bricked devices.
The cause of this issue is currently unknown. Hence rollback any upstream changes that may have causes this issue until the cause has been found.
The fixes commits are from around late May (24st) as the first reports of bricked devices was around early June.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
Setting IFS globally causes problems as mentioned by voidanix in the
Matrix chat. Specifically, it messes up splitting of words and
filenames for subsequent runs of the script. This avoids that and was
tested with both bash and zsh.
Signed-off-by: Ed Beroset <beroset@ieee.org>
This adds an explicit branch name to fix#250 in which the update
subcommand fails to work if the user has a fork of the main repository.
Signed-off-by: Ed Beroset <beroset@ieee.org>
Previously one could provide a new platform to the `devices` variable. This implies a new layer under `src/meta-smartwatch/meta-*`.
But the build configuration (bblayers.conf) is never updated with this new information as that file may already exists.
This adds support for updating the layer configuration whenever a new layer is added to the `layer_conf` variable and or a new layer becomes available in the `meta-smartwatch` repository.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
Since 376fb5ff5f most all of our ports fail to compile as they use a specific SRCREV for libhybris.
Eventually all ports should use the same SRCREV instead of something custom.
This works around the current issue by rolling back `meta-smartphone` until all ports use the same version of libhybris.
This provides a solution to https://github.com/AsteroidOS/asteroid/issues/238.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
When working on AsteroidOS recipes, it's convenient to have both a fork
of the subproject and also to set the upstream url. However, the
prepare-build.sh file would then only retrieve updates from the fork
rather than upstream. This addresses that issue by checking whether
there is an upstream or not; if not, it does the fork as it always had.
If there is an upstream, it pulls the named head from that fork instead,
making it easy and reliable to maintain one's own fork for contributions
back to the main project.
Signed-off-by: Ed Beroset <beroset@ieee.org>
Change the logic such that all layers are stored in an array so that we can always refer back to it.
This is used to update the layers and to make sure that they are on the correct branch.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
The SailfishOS related remote changes have been merged to the honister branch.
It's safe now to continue using the branch again.
Signed-off-by: Darrel Griët <dgriet@gmail.com>
git.merproject.org is obsolete all repos have moved to github.com/sailfishos.
These changes have been merged in meta-qt5 but only on the master branch.
Also see https://github.com/AsteroidOS/meta-asteroid/issues/66.
As of bitbake 7dcf317cc1 a new override syntax is used.
Not all current layers have switched to this new syntax yet (most notable meta-qt5), others have applied this new syntax in the upcoming Yocto Honister release instead of the current Hardknott release.
Due to the mixed usage of the different layers keep the current syntax and reconsider when moving our current build to the new Honister release.
