| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
| |
OE-core has renamed this recipe, and attempts to add a PROVIDES
compatibility item have been rejected.
https://lore.kernel.org/openembedded-core/CANx9H-B4-tEGjC_y0GiFrAeR24Z2PjocVX-=6CRbGBQ+rXytgg@mail.gmail.com/
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
|
|
|
|
|
|
|
|
| |
OE-core has renamed the bmap-tools forcing every downstream layer to
update their metadata. Thus it is not possible to actually be
compatible with both nanbield and scarthgap, so drop nanbield.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
|
|
|
|
|
|
|
|
|
|
|
|
| |
OE-Core updated compatibility in upstream commit 0e42326dfd6b
("layer.conf: Prepare for release, drop nanbield LAYERSERIES") to upcoming
Release 5.0 ("scarthgap") [1].
Update layer compatibility to match OE-Core, and drop "mickledore" as
according to [1] it is indicated as EOL.
Link: [1]: https://wiki.yoctoproject.org/wiki/Releases
Signed-off-by: Andrey Zhizhikin <andrey.z@gmail.com>
|
|
|
|
|
|
|
| |
If not specified "install" installs the files with rwxr-xr-x
permissions. Set the mode explicit to fix this.
Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
The recipe is part of meta-oe since [1]. It's available on master and
nanbield.
[1] https://git.openembedded.org/meta-openembedded/commit/meta-oe/recipes-devtools/microcom?id=2aa8abc9171aac23df050116d82d41eafb403094
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
| |
A future commit will drop the recipe as it's now part of meta-oe.
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
This update is based on the equivalent update commits in meta-rauc
from 2021.03.0 to 2023.08.0[1] and from 2023.08.0 to 2023.11.0[2].
[1]: https://github.com/rauc/meta-rauc/commit/6c5faed98b7124e1f084280b0981da349f38c8d6
[2]: https://github.com/rauc/meta-rauc/commit/b9938b796549a27bc481bac668d7104e58265bec
Signed-off-by: Leonard Göhrs <l.goehrs@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
poky/oe-core just released 'nanbield' and thus removed 'mickledore'
compatibility.
While at it, remove 'langdale' which is out of support.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
| |
There is no advantage in using PTX_ variables in a distro that is not
meant to be generic. It only adds unnecessary complexity.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
Port meta-yocto's commit [1] to harmonize with Poky.
[1] 66525f278143 ("poky: don't use space with +=")
https://git.yoctoproject.org/meta-yocto/commit/?id=66525f278143b7fe6082c8f5956cdff64b16ff58
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Generalize needed functionality to automatically convert images to be displayed
by platsch into a dedicated class. Currently only the RGB565 format is supported.
A recipe consuming that class could look as simple as e.g.
SRC_URI = " \
file://splash-1280x800.png \
file://splash-800x480.png \
"
inherit platsch
SPLASH_IMAGES = " \
splash-1280x800.png \
splash-800x480.png \
"
SPLASH_FORMAT = "RGB565"
and would produce a package containing the raw bitmaps in RGB565 format in the
default location ${datadir}/platsch (which normally is /usr/share/platsch) as
/usr
└── share
└── platsch
├── splash-1280x800-RGB565.bin
└── splash-800x480-RGB565.bin
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Starting with bitbake-2.2 which is part of yocto-4.1 ("langdale") classes can
explicitly be classified to be designed for global usage in the overall
configuration space or recipe-wise usage by moving them into the directories
"classes-global" or "classes-recipe" respectively (see Bitbake's commit [1] /
Poky's commit [2]).
Let's do this for meta-ptx' existing classes.
[1] f33ce7e742f4 ("BBHandler/cooker: Implement recipe and global classes")
[2] 7bd328f9d24b ("bitbake: BBHandler/cooker: Implement recipe and global classes")
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
While genimage itself is MIT-licensed, images created with genimage using
the 'inherit genimage' should not be forced to be MIT-licensed. The
standard for bbclass files which set LICENSE is to set it with '?=' so
the recipe can override it. Otherwise (with '=') it isn't possible for
the recipe to override it.
Signed-off-by: David Norris <david.norris@blueberrydaq.com>
|
|\
| |
| | |
conf: layer: add 'mickledore' to LAYERSERIES_COMPAT
|
|/
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
| |
ipk package management is more light-weight than rpm and sufficient for
most use cases. It requires less crude work arounds like package version
translations due to character restrictions on Yocto's side.
This should be taken as a recommendation, since projects are encouraged
to create their own distro (maybe copied from here) and PACKAGE_CLASSES
is set in local.conf(.sample) by default.
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
| |
platsch is a simple splash screen application meant to be run as pid 1.
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
|
| |
A future commit will rely on langdale, so drop kirkstone compatibility
now.
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
Among other features this new release brings (according to the git log):
- F2FS support
- Firmware Image Package (FIP) support
Signed-off-by: Leonard Göhrs <l.goehrs@pengutronix.de>
|
|
|
|
|
|
| |
Keep compatibility to 'kirkstone' first of all.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
| |
In poky commit 4f4a5dc9 ("lzop: remove recipe from oe-core") lzop recipe
and all dependencies on it were removed from oe-core because lzop is
seen as deprecated and hasn't seen any release since 2017.
Thus remove dependency on it from barebox build, too.
This will potentially break use cases where lzop is still used for
compression. Consider switching to different compression method then.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We currently expand only @IMAGE@ to ${GENIMAGE_IMAGE_FULLNAME}. Users
may want to expand arbitrary bitbake variables, e.g. ${MACHINE} or
self-defined ones. Add support for this and make @IMAGE@ a special case
of this. We don't do this for arbitrary variables to make variable
dependency generation easier. As users have to specify
GENIMAGE_VARIABLES explicitly, we don't need to generate vardeps in
python code (which would have to happen at recipe parse time, while
we only parse the files at do_genimage time).
We intentionally keep the @VAR@ syntax to avoid ambiguity:
exec-pre and exec-post commands have shell expansion to access genimage
provided environment variables like $IMAGEOUTFILE, so it would be
error-prone to reuse the same syntax for genimage.bbclass.
Preprocessing happens in a dedicated do_genimage_preprocess to allow
writing it in python while keeping do_genimage in shell. Doing it this
way instead of adding an intermediate python do_genimage yields better
error messages: logs from do_genimage are printed on task failure
instead of having just a python backtrace.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A genimage config may generate multiple images, either because multiple
image { } clauses are specified or because an exec-post generated
further images (e.g. image { exec-post = "gzip -k $IMAGEOUTFILE" })
We already deploy these extra images, but currently symlinks are only
created for the plain image and for the bmap file. Make this more
generic, so fixed symlinks are created at least for all images
using @IMAGE@ as part of their filename.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
|
|
|
|
|
|
|
|
| |
genimage.bbclass hardcodes the name of the genimage config. Users may
want to have multiple configs and select between them, so add a new
variable to make this straight forward to do.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
|
|
|
|
|
|
|
| |
Having the file name in the header can be useful when grepping, but it
was missing an 's', so fix that.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
OE-Core started in [1] to warn when it finds obsolete license identifiers as it
was decided to standardise on SPDX license names. Use the conversion script
introduced in OE-Core's commit [2] to silence the warnings.
[1] 9379f80f484f ("license/insane: Show warning for obsolete license usage")
[2] 512cd4ca91bc ("scripts: Add a conversion script to use SPDX license names")
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
|
|
| |
This makes it easier to find the upstream repository.
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Ryan Meulenkamp <Ryan.Meulenkamp@Nedap.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It is currently somewhat awkward to use include() from within
genimage.config, since one has to arrange for all the fragments to be
copied to ${B}; it is not enough to simply have them listed in
SRC_URI. So currently, we have this in our image recipe:
do_genimage:prepend() {
cp ${S}/*.config ${B}
}
do_genimage:append() {
# do_install takes everything in $B and puts it in $DEPLOYDIR. The
# .config fragments serve no purpose there.
rm -f ${B}/*.config
}
While the first is somewhat tolerable, the need for the second
fragment is rather annoying.
Now that genimage (since v15, and built against new enough libconfuse
which we do) has an --includepath option, add that to the command line
and point it at ${S} aka ${WORKDIR}, making include() work
out-of-the-box.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
| |
There are quite a few features and fixes between v14 and v15, and
while I can carry my own recipe for a newer genimage, let's bump the
recipe in meta-ptx.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
|
|
|
|
|
|
|
|
|
|
| |
It's useful to be able to pass extra custom options to genimage,
e.g. when one is building a newer genimage than the one defined in
meta-ptx to gain access to new features. For example, I wish to make
use of both --includepath (which is currently in genimage master) and
the just-proposed --configdump option.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The idea to obtain the exact kernel version looked simple:
* Depend on do_shared_workdir task of kernel to have its source dir
populated
* run get_kernelversion_file() method to obtain the exact version string
This is fine as it allows to set the 'version' field and let the 'linux'
field point to the exact kernel version file.
However, reality hits you when attempting to build an eSDK.
In an eSDK, we have a set of fixed signatures and shared state artifacts
for each of them. However, do_shared_workdir is different as it is a
setscene-whitelisted task that always gets executed in dry run mode to
trigger depending tasks.
This means, despite it has a sort of setscene task, this task has no
data associated.
Now when using the bootspec class in an image that we build an eSDK
from, we make do_shared_workdir an actual build dependency for the
do_rootfs task. This will result in bitbake trying to find and unpack an
sstate artifacts for it, which must fail.
In consequence, bitbake attempts to execute all the tasks upstream.
But, as we are in an eSDK build incapable of executing actual tasks, this
ends in an error message like:
| ERROR: Task linux-yocto.do_fetch attempted to execute unexpectedly
By not knowing any good alternative to obtain the exact kernel version,
this patch now simply removes the kernel version magic and thus the
dependency on do_shared_workdir.
For the 'version' field we default to ${PN} instead and the 'linux'
field now simply points to the '/boot/${KERNEL_IMAGETYPE}' symlink that
should exist on all systems.
In this form it does not allow differentiating multiple kernel in the
same directory, but this was never a real core skill of this class,
anyway.
The main purpose is still to have multiple supported dtbs in the same
rootfs.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As pkgconfig is potentially nedded to build some of barebox' host tools (e.g.
mxsimage [1] or imx-usb-loader [2]) starting with Yocto's commit [3] we have to
explicitely depend on it, otherwise the information that should have been
gathered via 'pkg-config [...]' is missing with the result that building those
tools breaks.
[1] https://git.pengutronix.de/cgit/barebox/tree/scripts/Makefile?h=v2021.11.0#n32
[2] https://git.pengutronix.de/cgit/barebox/tree/scripts/imx/Makefile?h=v2021.11.0#n4
[3] 04b37f914ed1 ("layer.conf: Extend recipes not to install without explict dependencies")
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
The content was moved to base.bbclass.
This adapts poky commit 9d12605d:
poky: Reproducible builds are now the default
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adapts the original poky commit b0fd19e5:
local.conf.sample: Bump version so users update their config
It has been a while since we've done this but with the new syntax, users
need to update their local.conf if based off the template. Update the
versions to trigger this.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
As promised in [1] clean up TARGETCFLAGS now as they are of no use anymore for
current barebox releases (starting with v2020.08.0).
[1] https://github.com/pengutronix/meta-ptx/pull/74#issuecomment-883540834
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
|
|
|
|
|
|
|
| |
do_task[depends] always expects dependencies in the form of
<recipe>:<task>. This we missed for virtual/bootloader.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Done with conversion script:
scripts/contrib/convert-overrides.py meta-ptx
Update layer compatibility to honister.
Note: New override-style syntax is supported by latest version of poky
hardknott. Thus compatiblitiy should actually be given.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Barebox' commit [1] contained in v2020.08.0 for the first time broke building
the userspace programs for the target system (bareboxenv & Co.). Fix that by
offering the needed compiler flags not only in the present parameter
"TARGETCFLAGS", but additionally in the newly introduced one "userccflags".
Note that this might not be a definitive solution as it's unclear whether
userccflags is meant as user API in Linux (which barebox Kbuild is synchronized
with).
The same issue should happen with kernel samples, which are also target tools.
Those seem not to be handled by poky. Selftests are supported, but those have
their own Makefile and don't result as part of the normal Kbuild procedure.
[1] c008d836d8f4 ("scripts: use 'userprogs' to build programs for target")
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de>
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If BAREBOX_FIRMWARE_DIR is set, barebox' CONFIG_EXTRA_FIRMWARE_DIR is
set to its value. Note that the original value of
CONFIG_EXTRA_FIRMWARE_DIR is overwritten in that case.
This allows setting the firmware search path directly without the need
of copying firmware files around, e.g.:
BAREBOX_FIRMWARE_DIR = "${STAGING_DIR_TARGET}/firmware"
Currently (v2021.06.0) the imxcfg directive "signed_hdmi_firmware" does
not consider CONFIG_EXTRA_FIRMWARE_DIR, so setting BAREBOX_FIRMWARE_DIR
does not affect that. A possible workaround is:
BAREBOX_FIRMWARE_DIR = "${B}/firmware"
do_compile_prepend() {
mkdir -p ${BAREBOX_FIRMWARE_DIR}
cp ${DEPLOY_DIR_IMAGE}/signed_hdmi_imx8m.bin ${BAREBOX_FIRMWARE_DIR}
}
This works because "signed_hdmi_firmware" also expects the firmware in
`${B}/firmware`. Note that other needed firmware files need to be copied
to the same destination.
Suggested-by: Enrico Jörns <ejo@pengutronix.de>
Signed-off-by: Bastian Krause <bst@pengutronix.de>
|