The build logic in source/cmake/FindNeon.cmake caused the x265 build
system to always think that the CPU supports neon: it was looking in
/proc/cpuinfo, which of course is wrong when cross-compilation, but
then the sequence of grep was interacting badly with CMake, causing
the build system to always conclude that the CPU supports NEON.
This causes runtime issues on ARMv6.
Setting -DCROSS_COMPILE_ARM=1 fixes this, as it tells the x265 build
system we are cross-compiling and it skips its bogus NEON check. So
for ARMv6, we pass -DCROSS_COMPILE_ARM=1.
But then, we still want NEON for ARMv7 processors with NEON, so this
commit adds a patch that allows to explicitly specify whether the CPU
supports NEON, in the -DCROSS_COMPILE_ARM=1 case, and we use this
option when BR2_ARM_CPU_HAS_NEON.
For those wondering why -DCROSS_COMPILE_ARM=1 is not passed for all
ARM platforms: it's because from the perspective of x265, only ARM >=
v6 is ARM: it has assembly code that needs at least ARMv6. Earlier ARM
platforms are not detected as ARM by the x265 build logic, and
therefore fallback on generic code.
This has been build-tested on:
- ARMv5: generic code is used, no assembly
- ARMv6: assembly code is used, but not with NEON support
- ARMv7 with NEON: assembly code is used, with NEON support
Reported-by: David Barbion <davidb@230ruedubac.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 1fbd26f831)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
As we're about to do some other fixes in x265 ARM build, let's replace
the patch fixing gcc option by:
- A patch generated by git format-patch now that x265 upstream uses
Git
- Is re-submitted upstream
- Only drops the flags, and does not add bogus ARM architecture
aliases that don't exist in the CMake world.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9d62687d00)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
GNU Octave supports the --with-sndfile configure option since v4.0.0.
For reference, commit 40ea68b4b2 "package/octave: new package"
introduced the package at v7.1.0.
This commits adds this optional support.
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit a5ce48fe45)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This allows uvw to be again in sync with the libuv package, following
the bump of libuv to v1.48 in commit
bd2f99246c ("package/libuv: security
bump to version 1.48")
The hash of the license file has changed due to a copyright year
update.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit c54134bf69)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When build host has a many CPUs (more that 20) and GNU Make 4.4
(included for example in Fedora 40), fwts can randomly fail to build.
This commit adds a package patch to fix the issue.
Fixes:
mv: cannot stat 'dtcompilerparser.tab.c': No such file or directory
mv: cannot stat 'prparser.tab.c': No such file or directory
See also:
https://github.com/fwts/fwts/issues/7
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit f6d9fa3692)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add bsdunzip option to avoid the following build failure raised since
bump to version 3.7.1 in commit 7991d2c48a and
c157e4ce8e:
CCLD bsdunzip
/home/autobuild/autobuild/instance-4/output-1/host/lib/gcc/powerpc-buildroot-linux-musl/12.3.0/../../../../powerpc-buildroot-linux-musl/bin/ld: cannot find -lz: No such file or directory
By handling it explicitly, we can specify whether the dynamic or static
library should be used, according to what we actually have available.
Fixes:
- http://autobuild.buildroot.org/results/aaf7fdefa0b7bd7e5ac743487c197544c1a2ce6f
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Tested-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit bbc08b15ae)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This issue was introduced back in 2009 with commit
e11fe847b2 that created the generic
package infrastructure.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit bd0d5498ea)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When BR2_SHARED_LIBS=y (shared libraries only) and BR2_PIC_PIE=y
(Build code with PIC/PIE), the toolchain-wrapper will try to enable
position-independent code/executables. See [1]. This configuration
is a common default.
Xvisor was likely tested only with RISC-V gcc from [2], which will
not enable PIE by default. Since Xvisor is a Type 1 hypervisor, it
needs the same kind of special treatment as U-Boot or Kernel.
This commit adds a patch to explicitly force static linking and
disable PIE for RISC-V architecture.
[1] https://git.buildroot.org/buildroot/tree/toolchain/toolchain-wrapper.c?h=2023.05.1#n392
[2] https://github.com/riscv-collab/riscv-gnu-toolchain
Signed-off-by: Julien Olivain <ju.o@free.fr>
[Arnout: renumber patch]
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 28925c6c86)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Integrate a patch, which was submitted upstream, to work-around an
issue with the SPARC kernel headers. The kernel headers have been
fixed in the upstream kernel (in the to-be-released 6.10 kernel), but
we need a workaround for the older toolchains, until we believe 6.10
is "old" enough to be able to drop the workaround.
Fixes:
http://autobuild.buildroot.net/results/272c464ed4f9392535fa3b7613218dbd03acf901/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit b162aab7c6)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Although the asciidoc toolchain accepts any number of ~ to delimit a
listing block (i.e. a code block), it is actually specified to be
exactly four, i.e. ~~~~. Currently, a mix of diffrent numbers of ~ are
being used - sometimes even a different number at the beginning and at
the end of the block.
Normalize this to always use exactly four ~ for the delimiter.
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 447fa1fca4)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Currently the text for each package infra that mentions the usage of
variables already provided by the generic infra diverge from each other:
- some (golang, kconfig, python) add a cross-referece to the generic
infra chapter;
- kconfig does not list any example;
- some mention _LICENSE as an example, others don't;
- some (cargo, golang, python) add an 'etc.' at the end of the examples,
giving the idea that can be more symbols provided by the generic
infra than the ones listed;
- most have the text 'works by defining a number of variables before
calling the +<macro-name>+ macro', except golang and kconfig;
- some actually list 'A few additional variables' but keep using some
old reference as 'An additional variable';
- some say 'First, all the package metadata' and other only 'All the
package metadata';
- most mention _SUBDIR as an example of variable supported by the
generic infra, even the generic infra manual not mentioning it.
Improve the correctness for the manual by standardizing the text among
the package infras:
- use the same text "All the package metadata information variables that
exist in the generic package infrastructure also exist in the
<name> infrastructure:" for all of them;
- add the cross-reference for all of them;
- remove the examples of variables inherited from the generic infra -
this also solves the _SUBDIR problem, there no longer is any reference
to _SUBDIR;
- wrap the modified text at 80 columns;
- add "macro" to golang and luarocks infra;
- use "A few additional variables" for qmake and waf.
At same time, add a missing format on golang manual for
BR2_PACKAGE_HOST_GO_HOST_ARCH_SUPPORTS.
Cc: Eric Le Bihan <eric.le.bihan.dev@free.fr>
Cc: Thomas De Schampheleire <thomas.de_schampheleire@nokia.com>
Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
[Arnout:
- remove the examples;
- add "the" where "macro" was added;
- rewrite the preceding paragraphs for kconfig to make it more
consistent.
]
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 4286c89f9d)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Each linux tool uses a fragment of a .mk file, named, for instance:
package/linux-tools/linux-tool-cpupower.mk.in
So currently check-package does not check these files.
Add the support in check-package script.
At the same time, factor out a function to derive package prefix from
the filename being checked, so the fix (calling os.path.splitext twice)
can be applied in a single place.
Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
[Arnout: add docstring and explain double splitext to
get_package_prefix_from_filename]
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit ccc12c0f24)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fix the following musl build failure with gcc 14:
daemon/old_main.c: In function 'old_main':
daemon/old_main.c:56:11: error: implicit declaration of function 'strcmp' [-Wimplicit-function-declaration]
56 | if(!strcmp((which_mouse->opt_dev),"-")) fd=0; /* use stdin */
| ^~~~~~
daemon/old_main.c:35:1: note: include '<string.h>' or provide a declaration of 'strcmp'
34 | #include "headers/gpmInt.h" /* daemon internals */
+++ |+#include <string.h>
35 |
While at it, add Upstream tags to other patches
Fixes:
- http://autobuild.buildroot.org/results/3192740863d57e72e90e622683e7b9fcc057fa8b
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 50b18ed68c)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit 8f88a644ed ("support/scripts/apply-patches.sh: set the maximum
fuzz factor to 0") reduced the fuzz factor.
Due to this change, gpm fails to build with output:
Applying 0003-src-Makefile.in-Really-install-unversioned-solibrary.patch using patch:
patching file src/Makefile.in
Hunk #1 FAILED at 115.
1 out of 1 hunk FAILED -- saving rejects to file src/Makefile.in.rej
This commit refreshes the package patch on the current package version.
While we're at it, change to use a proper Upstream: tag and remove from
.checkpackageignore.
Fixes: 8f88a644ed
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit d47ae17adf)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Release Notes: https://nodejs.org/en/blog/release/v20.12.0
LICENSE hash changed due to changes in vendored components:
* copyright year updates [1][2]
* version name update [3]
* change from "Unicode, Inc. License - Data Files and Software"
to the "Unicode License v3" [4]. This is in the icu vendored
dependency which is unused in buildroot.
* new build tooling script [5] under BSD style license
NodeJS tests are passing:
$ ./utils/docker-run ./support/testing/run-tests -o ./outputs/ -k tests.package.test_nodejs
13:40:05 TestNodeJSModuleHostSrc Starting
13:40:06 TestNodeJSModuleHostSrc Building
14:05:52 TestNodeJSModuleHostSrc Building done
14:06:02 TestNodeJSModuleHostSrc Cleaning up
.14:06:02 TestNodeJSModuleHostBin Starting
14:06:03 TestNodeJSModuleHostBin Building
14:24:25 TestNodeJSModuleHostBin Building done
14:24:31 TestNodeJSModuleHostBin Cleaning up
.14:24:31 TestNodeJSBasic Starting
14:24:32 TestNodeJSBasic Building
14:42:53 TestNodeJSBasic Building done
14:43:02 TestNodeJSBasic Cleaning up
.
----------------------------------------------------------------------
Ran 3 tests in 3776.679s
OK
[1] 347e1dd06a
[2] b88170d602
[3] c8233912e9
[4] 625fd69b76
[5] b5bc597871
Signed-off-by: Marcus Hoffmann <buildroot@bubu1.eu>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit e86fa6bdb4)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When packaging data_files in a wheel, those files shall have the destination
path expressed as relative to the wheel installation.
The reason for that is the wheel installation shall be contained inside a
specific folder and/or in site-packages.
When building the wheel file by the bdist_wheel command, python will package
files with absolute paths as python files ending up installing them in
site-packages.
Fixes:
http://autobuild.buildroot.net/results/36ac5278d19195a21c3d02d087965e08f49228ef
Signed-off-by: Angelo Compagnucci <angelo.compagnucci@gmail.com>
Tested-by: Fiona Klute <fiona.klute+wiwa@gmx.de>
[Arnout: add Upstream tag to the patch]
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit c90b87156d)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When installing a wheel in pep517 compatibility mode, the pyinstaller.py is
invoked with the --data parameter which point to the directory in which data
files should be installed.
Actually the --data parameter points to the /usr subdirectory which is indeed
wrong cause it shall point to the root directory where the wheel will be
installed.
This fixes the problem of having configuration files installed in /usr/etc
instead of /etc.
Signed-off-by: Angelo Compagnucci <angelo.compagnucci@gmail.com>
Tested-by: Fiona Klute <fiona.klute+wiwa@gmx.de>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 98f452f974)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since its inception in 146498d13c (package/bcc: new package), not all
the dependencies of bcc are mirrored in the comment to be displayed when
those dependencies are not met.
Fix that by adding the missing pieces:
- wchar,
- threads
- dynamic libs,
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Jugurtha BELKALEM <jugurtha.belkalem@smile.fr>
Cc: Romain Naour <romain.naour@gmail.com>
Cc: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 76051e6531)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes http://autobuild.buildroot.net/results/1d5804974153f96e80f5200793ec8cd12c9fa18a/
GCC 14 complains if the arguments to calloc (nmemb, size) are swapped
around, causing a build failure as dump1090 builds with -Werror:
net_io.c:107:34: error: 'calloc' sizes specified with 'sizeof' in the earlier argument and not in the later argument [-Werror=calloc-transposed-args]
107 | if (!(service = calloc(sizeof(*service), 1))) {
Add a patch from an upstream pull request to fix that.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit fd358b97c0)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commits 3988f2d923 and
0d91281d7b added hashes for the chromebook
defconfigs, but forgot to remove the exception from .checkpackageignore.
Regenerate .checkpackageignore, which removes those exceptions.
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 5e36efd7ac)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The patch adds .hash files for Arm trusted firmware, Linux, Linux headers
and U-Boot and then enables BR2_DOWNLOAD_FORCE_CHECK_HASHES. With this, we
can now drop the defconfig from .checkpackageignore.
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 751228436a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
qt5webkit build currently breaks on the following error:
XSLStyleSheetLibxslt.cpp:148:129: error: invalid conversion from ‘void (*)(void*, xmlError*)’ {aka ‘void (*)(void*, _xmlError*)’} to ‘xmlStructuredErrorFunc’ {aka ‘void (*)(void*, const _xmlError*)’} [-fpermissive]
148 | XMLDocumentParserScope scope(cachedResourceLoader(), XSLTProcessor::genericErrorFunc, XSLTProcessor::parseErrorFunc, console);
| ^
| |
| void (*)(void*, xmlError*) {aka void (*)(void*, _xmlError*)}
This error is due to an API update in libxml2, enforcing const on more
struct in version 2.12.0 (see [1]). Buildroot now tracks v2.12.5.
Upstream Webkit project has already issued the corresponding fix ([2]),
which updates corresponding internal prototypes depending on libxml2
version, but the qt5webkit version tracked in buildroot does not integrate
the corresponding Webkit version.
Fix this build issue by bringing the upstream patch "as is" from Webkit
No autobuilder references because this build error was hidden by
another build error fixed in a previous patch.
[1] https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.12.0
[2] 1bad176b24
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit e24a117388)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
qt5webkit build currently fails with the following error:
[...] parser.rb:587:in `block in parseSequence': undefined method `=~' for an instance of Annotation (NoMethodError)
from <internal:kernel>:187:in `loop'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:586:in `parseSequence'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:654:in `block in parseSequence'
from <internal:kernel>:187:in `loop'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:586:in `parseSequence'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:814:in `parseData'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:818:in `parse'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:780:in `block in parseSequence'
from <internal:kernel>:187:in `loop'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:586:in `parseSequence'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:625:in `block in parseSequence'
from <internal:kernel>:187:in `loop'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:586:in `parseSequence'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:814:in `parseData'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/parser.rb:818:in `parse'
from /home/alexis/src/buildroot_min/output/build/qt5webkit-5.212.0-alpha4/Source/JavaScriptCore/offlineasm/generate_offset_extractor.rb:68:in `<main>'
This issue is due to =~ being marked as deprecated since a few Ruby
versions, and finally removed in 3.2.0 [1]. This now breaks the build since
buildroot has moved to Ruby v3.3.0.
The corresponding fix has already been issued in upstream Webkit project
[2], but qt5webkit version tracked in buildroot does not have the
corresponding webkit version pulled. Fix this build error by bringing the upstream
patch. The patch is slightly modified (exclude part about Changelog file,
which is absent from qt5webkit)
Fixes:
http://autobuild.buildroot.net/results/21397b110fe02e5711ecb1d35be2108221751b0a/
[1] https://www.ruby-lang.org/en/news/2022/12/25/ruby-3-2-0-released/
[2] c7d19a492d
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 51ef6691ac)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit 8f88a644ed ("support/scripts/apply-patches.sh: set the maximum
fuzz factor to 0") reduced the fuzz factor.
Due to this change, qt5webkit fails to build with output:
Applying 0004-Remove-invalid-g_object-declarations-to-fix-build-wi.patch using patch:
patching file Source/WTF/wtf/glib/GRefPtr.h
Hunk #1 FAILED at 29.
1 out of 1 hunk FAILED -- saving rejects to file Source/WTF/wtf/glib/GRefPtr.h.rej
This commit refreshes the package patch on the current package version.
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 6d0cce7dab)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since there are no more packages using FTP-hosted files, we can drop
the --passive-ftp option from genrandconfig, as it would cause
problems on systems that use wget2.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 881a8f1346)
[Peter: slightly reword commit message, no curl backend here]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
We're going to add hashes soon, so we'll need to have that directory
populated with hash files, and it would then be a bit confusing to not
have the patch file in the patches directory...
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Alex Suykov <alex.suykov@gmail.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 552a8cec8c)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Change download location as the new version is not available at the
old location.
Remove already upstreamed patch.
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 0deb8c4c5d)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Backport a patch from upstream that adds the magic number for the
bcachefs superblock. Otherwise, systemd 254.13 fails to compile with
the latest kernel versions.
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit f3c1f667dc)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
