| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Setting the new variable to 'YES' allows to inject into <PKG>_BUILD_DIR
instead of <PKG>_DIR as before.
When building out-of-tree some bootloaders like U-Boot expect injected
files in the build dir, not in the source dir. Backwards compatibility
is ensured, the source dir is still the default, the new inject dest is
optional and in case set it overwrites the default.
For using this in u-boot package, on top of the things already done to
use the old inject mechanism in general, add this line to
'rules/u-boot.make' for example:
U_BOOT_INJECT_OOT := YES
Signed-off-by: Alexander Dahl <ada@thorsis.com>
Acked-by: Michael Riesch <michael.riesch@wolfvision.net>
Message-Id: <20240506105051.323546-5-ada@thorsis.com>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
pkg_source was defined as "$($(1)_DIR)" which is the same as pkg_dir in
ptxd_make_world_common. We can use pkg_dir directly. Add a safe-guard
to bail out early if that var is empty.
Signed-off-by: Alexander Dahl <ada@thorsis.com>
Acked-by: Michael Riesch <michael.riesch@wolfvision.net>
Message-Id: <20240506105051.323546-3-ada@thorsis.com>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
One line above $target is set to "${var}/$(subshell)" and even if $var
_and_ the result of the subshell call is empty, there's still the slash.
Thus target is never empty.
Signed-off-by: Alexander Dahl <ada@thorsis.com>
Acked-by: Michael Riesch <michael.riesch@wolfvision.net>
Message-Id: <20240506105051.323546-2-ada@thorsis.com>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
This is necessary for nested "menuconfig" etc.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
Updating the stamp here is done so that updated to the .config timestamp
done by the migrate helper will not always cause the config to be saved.
However, if the migrate helper actually changes something in the file
then don't update the timestamp to ensure that the config is actually
updated with those changes.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If the group is explicitly overwritten and not used from the directory
that is copied, then it does not make sense to keep the set-group-ID bit
meant for a different group.
This is relevant for install_alternative_tree. If the directory where a
BSP is cloned has the set-group-ID bit set, then the whole source tree
inherits it, including all directories in projectroot/. As a result
install_alternative_tree would copy it.
In such a use-case install_alternative_tree is typically called with
group=0 (or another explicitly specified group) so that can be used as
indicator when the set-group-ID bit should be dropped.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
npm can create an sbom from the package-lock.json. Do that and reference
the result in the full-bsp-report.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
of the source archive
This is especially useful in combination with <PKG>_STRIP_LEVEL when
only a subdirectory is relevant.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
Some packages contain vendored dependencies as a tarball. With this,
<PKG>_PARTS can be used for those and the default extract stage can
extract them.
Extracting early instead of allowing the package buildsystem to do this,
makes it possible to patch the vendored dependencies.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
| |
This step reads the Cargo.lock file. For packages where this is not
provided by upstream, it will be created with a patch, so do this step
when the patches are applied.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
This is basically the same thing already done for host-* packages. It
allows those packages to share the version etc. from the host or target
package.
Remove the now obsolte definitions from the rule files.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
The downloads actually happen in the source archive target. Here
pkg_stage is not set so the fifo to limit parallelization does not
exist.
Set pkg_stage if needed to fix this.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
Put stderr directly into the logfile and let wget output something
useful to help with debugging.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To handle all downloads manually, PTXdist basically fakes vendoring all
dependencies. This can cause problems with workspaces where packages
inherit from the workspace. 'cargo vendor' used to have the same
problem[1] and fixed it by rewriting the Cargo.toml.
Do the same thing here to fix this.
[1] https://github.com/rust-lang/cargo/issues/11192
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
counterpart
In that case, "prompt" and "help" lines were added to the .in file and
"SECTION" was set to "project_specific" instead of "hosttools_noprompt".
Signed-off-by: Markus Heidelberg <m.heidelberg@cab.de>
Message-Id: <20240304072708.50708-1-m.heidelberg@cab.de>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
ptxd_get_ptxconf needs the kconfig symbol with the PTXCONF_ prefix.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
Skip during urlcheck and balance the serialize fifo.
Otherwise a token is added that was never removed.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
Sometimes users rename the downloaded file to '<filename>.somthing' to
test the download again. Then they are confused that the original file
is gone.
So restrict the pattern a bit and try to delete only files that where
created during a previous aborted download.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
Other build-ids by be shorted, e.g. those generated with the 'fast'
style from lld. Fallback to the old debug file path in that case.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
Provide the options with the correct prefix for packages that need to
specify it explictly and use them as default when needed.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
details
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
<PKG>_SOURCES cannot be used to match back to the package parts because
the same source path may be used by multiple packages.
So define <PKG>_PARTS that contains <PKG> and can be extended with
<PKG>_EXTRA etc. for additional sources. In that case,
<PKG>_EXTRA_SOURCE, etc. is expected to be defined.
This makes it possible to extract all components in the default rule
among other things.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
gen_compile_commands.py defined the wrong directory when creating
compile_commands.json for external kernel modules: It must be the kernel
build tree instead of the module build dir.
To fix that, move the compile_commands.json generations into the core
and fix it afterwards.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
Create it manually if the build-system does not do this for us.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Multi platform kernels built by the kernel-fit rule before are not
usable in U-Boot, which adheres more strictly to the Flat Image Tree
(FIT) spec than barebox does. For image sub-type 'kernel' U-Boot
expects valid 'load' and 'entry' addresses. Setting those addresses to
a generic value is impossible however for generic kernels, e.g. the
DistroKit v7a kernel needs different addresses on at91 SAMA5 than
on i.MX6. barebox circumvents this by ignoring missing 'load' and
'entry' addresses and finding a suitable address by itself. For using
generic kernels in U-Boot you need sub-image type 'kernel_noload' and
dummy addresses, which is also supported in barebox from v2024.01.0
release onwards.
Link: https://github.com/open-source-firmware/flat-image-tree/releases/download/v0.8/fit-specification-v0.8.pdf
Link: https://lore.barebox.org/barebox/20231129203106.2417486-1-a.fatoum@pengutronix.de/
Signed-off-by: Alexander Dahl <ada@thorsis.com>
Message-Id: <20240202151209.2535721-3-ada@thorsis.com>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
| |
The python wrapper may be used to call host-python3. In that case, it
must be called as python3.x so fix the shebang accordingly.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some variables corresponding to ptxdist parameters use "true" as value
when given, others use "1". PTXDIST_DIRTY, PTXDIST_FORCE and
PTXDIST_PEDANTIC are in the first group according to
git grep -E 'PTXDIST_(DIRTY|FORCE|FEDANTIC)='
. Adapt ptxd_make_nested_ptxdist() accordingly.
Fixes: 5169ae99daf8 ("add support for nested PTXdist execution")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Message-Id: <20231110093705.45879-2-u.kleine-koenig@pengutronix.de>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
| |
If PTXDIST_WORKSPACE has two levels of symlink indirection then the
include path produced by cmake don't match the expected paths and are
filtered out by the toolchain wrappers.
Reproduce the cmake symlink handling and add the resulting paths to the
whitelist to fix this.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
| |
The --cross-file can be used multiple times and all files will be
combined. So don't expand the existing file.
Instead, write the extra options so individual files.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
| |
'ptxdist all*config ...' can be used to set all new symbols to a certain
value by setting KCONFIG_ALLCONFIG to the used ptxconfig. In that case,
run migrate to update the existing symbols.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
This section is for BSP packages, not for upstream.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
| |
Images and their reports where not always recreated correctly, e.g. when
a package was disabled.
Fix this by adding packages and files to the cfghash and depend on it
for the reports.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
| |
There may be not scripts. Don't call sed in that case. Otherwise it will
print a warning:
sed: no input files
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
| |
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
| |
Signed-off-by: Roland Hieber <rhi@pengutronix.de>
Message-Id: <20230912165759.3492269-2-rhi@pengutronix.de>
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
CMAKE_SYSTEM_PROCESSOR defaults to 'uname -m' and this is i*86 for
32-bit x86. So just use $PTXCONF_ARCH_STRING.
Without this building qt6 fails with:
In file included from .../qtmultimedia/src/3rdparty/pffft/pffft.c:141:
.../include/xmmintrin.h: In function 'passf2_ps':
.../include/xmmintrin.h:193:1: error: inlining failed in call to 'always_inline' '_mm_sub_ps': target specific option mismatch
.../qtmultimedia/src/3rdparty/pffft/pffft.c:148:19: note: called from here
.../qtmultimedia/src/3rdparty/pffft/pffft.c:290:23: note: in expansion of macro 'VSUB'
[...]
Because -DPFFFT_SIMD_DISABLE is not added to COMPILE_OPTIONS.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
|