0001-ext-opcache-config.m4-enable-opcache.patch is dropped.
This patch could be dropped because the new version now uses
AC_CACHE_CHECK, and we can just pass ac_cv_xxx to it instead of
using a local patch.
0008-ext-imap-config.m4-fix-include-paths.patch is dropped.
ext/imap has been removed from php in this new version. See
https://github.com/php/php-src/pull/13190. As a result of this
removal, the corresponding PACKAGECONFIG is removed from this new
version.
0001-Change-whether-to-inline-XXH3_hashLong_withSecret-to.patch is
dropped. It has been merged in this new version.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Drop 0003-php-remove-host-specific-info-from-header-file.patch.
Instead we export PHP_UNAME = "Linux" to achieve the same effect.
Drop 0002-build-php.m4-don-t-unset-cache-variables.patch.
The related ac_cv_lib_xxx and ac_cv_func_xxx settings in this recipe
are also removed. This patch is not needed from the my build testing
result.
Drop 0009-php-don-t-use-broken-wrapper-for-mkdir.patch.
This patch says that the wrapper is broken, but does not say why.
Without this patch, things still build.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Native build on aarch64 breaks with:
| {standard input}: Assembler messages:
| {standard input}:169: Error: selected processor does not support `xpaclri'
| {standard input}:411: Error: selected processor does not support `xpaclri'
| {standard input}:859: Error: selected processor does not support `xpaclri'
| {standard input}:1140: Error: selected processor does not support `xpaclri'
Signed-off-by: Alex Kiernan <alex.kiernan@gmail.com>
Signed-off-by: Alex Kiernan <alexk@a-squared-projects.uk>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Add a patch to not use AM_INIT_AUTOMAKE as automake isn't actually used,
and just let autotools.bbclass run autoreconf.
Signed-off-by: Ross Burton <ross.burton@arm.com>
As of "autotools: don't try and find in-tree macros" in core, acpaths is
no longer used.
Also switch to out-of-tree builds as this appear to work now.
Signed-off-by: Ross Burton <ross.burton@arm.com>
cpuid's Makefile passes -s to install to strip the binaries, but as this
uses the _host_ strip it can fail if that strip doesn't know about the
target (for example, building cpuid for x86 on aarch64):
install -D -s -m 755 cpuid .../image/usr/bin/cpuid
strip: Unable to recognise the format of the input file `.../image/usr/bin/cpuid'
Set INSTALL_STRIP='' so that the binaries are not stripped, and remove
the INSANE_SKIP for already-stripped.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Changelog:
==============
- New architecture: mips
- New architecture: loongarch
- Output buffering can now be unconditionally disabled
- New provider: profile
- Incorrect stack management when accessing tracepoint data
Signed-off-by: Wang Mingyu <wangmy@fujitsu.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The base problem this is trying to solve: capnproto has two main
components: libraries that can be linked against, and binary files
that can generate c++ code from capnproto definitions. When cross-compiling
one needs to use the cross-compiled libraries, but with the native
definition-compilers.
When linking against capnproto from another recipe, cross-compilation
fails, because the cmake files generated by capnproto verify the
existence of binary files in $bindir. When using a cross-compiled
version of capnproto, these binary files do not exist in RECIPE_SYSROOT, so
the compilation fails.
The previous patch: the previous solution patched the CMakeLists.txt
file in a way that it didn't export nor install the generated binary files.
Accidentally this also happened with native build (and happened knowingly with
target builds).
The new patch: instead of not installing and not exporting the binaries,
just install them without exporting, when creating a target build.
During compilation check is CMAKE_CROSSCOMPILING is set (coming from
cmake.bbclass) - if it is set, only install the binaries, without
exporting. When it is not set, resort to the original behavior.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
When LAPACKE is enabled, the QA check fails with the following error:
ERROR: lapack-3.12.1-r0 do_package_qa: QA Issue: lapack-ptest rdepends on lapack-dev [dev-deps]
ERROR: lapack-3.12.1-r0 do_package_qa: Fatal QA errors were found, failing task.
The root cause is that `lapack-ptest` accidentally installed `lapacke.pc`,
which is a development file intended to be included only in the `lapack-dev` package.
To resolve this: Remove `lapacke.pc` from the `lapack-ptest` installation
Signed-off-by: Zhang Peng <peng.zhang1.cn@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The correct dependency name for pyelftools is "python3-pyelftools".
Set the dependency name accordingly to avoid build failure with
this PACKAGECONFIG.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
As a build time dependency, python3 provides python3-core - set the
PACKAGECONFIG accordingly to avoid build failure.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
In case unwind PACKAGECONFIG is enabled, do_configure task
fails with the following error message:
| configure: error: Cannot enable both --with-libunwind and --with-elfutils
To avoid this error, disable elfutils with this PACKAGECONFIG.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The cmake files generated by this recipe contain a number
of interfacing libraries, with absolute paths, which also
contain the TMPDIR. The TMPDIR part of these paths were
just truncated by the recipe, to avoid the corresponding
QA check error.
However this made it impossible to to link against this
library from other recipes. When one tried, the find_package
command failed with the following error:
| Run Build Command(s): ninja -v -j 12 all
| ninja: error: '/usr/lib/libjsoncpp.so', needed by 'jsonrpctest', missing and no known rule to make it
To avoid this, instead of just truncating this path, set it to
${CMAKE_SYSROOT} placeholder (which is resolved by CMake to
the correct value at build-time, using the variable set by
cmake.bbclass)
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Changelog:
=============
* Mostly related to bazel build support.
* Min bazel supported is now 7 or higher, as WORKSPACE files are removed
* Minor C++ codegen fix removing extra semicolon
* Rust Full Reflection
* Mostly documentation updates hosted at https://flatbuffers.dev
Signed-off-by: Wang Mingyu <wangmy@fujitsu.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This reverts commits 4271cc2827 ("android-tools: Add flag to enable
adbd service (#147)") and b85ae59968 ("android-tools 10: Add flag to
enable adbd service"). These two commits added optional
ROOTFS_POSTPROCESS_COMMAND touching the flag file, however this variable
doesn't make any sense in the package recipe, it is only concerned in
the image context.
Drop useless code from the android-tools recipe.
Fixes: 4271cc2827 ("android-tools: Add flag to enable adbd service (#147)")
Fixes: b85ae59968 ("android-tools 10: Add flag to enable adbd service")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
protobuf-lite is a package, not recipe, so current code leads to
ERROR: Nothing PROVIDES 'protobuf-lite'
Moving it to 4th part (rdepends) is not needed as grpc links against it
so that rdependency is asutomatic.
Therefore protobuf-lite should be removed.
Note: for native, this packageconfig does not even compile and for
target it fails to link.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
googletest is needed only if we would build tests, and then it should be
conditional on PTEST_ENABLED
libnsl is not needed. there is no linking command in any cmake.
The only grep hit is in cares.cmake to disable it.
Commit 2ffc7b45d6 which added it talks
about host distros, but also such case would be wrong when adding it to
general depends (should be in native only depends) and it would have to
have be reference in cmake (in grpc or in one of recipe-sysroot).
Also sort the dependencies in the process.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
After the migration from Mercurial to GitHub the homepage has
changed and SIP is now licensed under the BSD-2-Clause license. The
upgrade to version 6.10.0 brings the following changes:
- SIP_ABI_VERSION has been added as a module attribute of the sip
module. This is an integer that represents the ABI version that
the module implements as a 3 part hexadecimal number.
This has been implemented in ABIs v12.17 and v13.10.
- The ability of sip-module to generate sdists for older ABI
versions has been restored. The oldest that can be generated are
v12.9 and v13.1 which are the oldest non-deprecated versions.
- The default minor ABI version to use is now the latest minor
version for a particular major version (rather than 0 as it was
previously).
- The --abi-version command line option has been added to sip-build,
sip-install and sip-wheel to specify the ABI version of the sip
module to generate code for.
- The %MinimumABIVersion directive is used to specify the minimum
sip module ABI to target in order to ensure that all calls to the
public API from handwritten code are available.
A consequence of this change is that passing string values to the
/Deprecated/ annotation no longer requires an ABI version that
supports them to be targeted. If the targeted ABI version does
not support them they are simply ignored.
- The value of a string annotation uses the ; and : characters to
allow platform- and feature-specific values to be defined. These
special characters can now be escaped using a leading \\. This
capability has now been documented.
- The license designator in setup.py for the latest versions of the
sip module are now correctly specified as BSD-2-Clause.
- The code generation has been fixed for classes created by
instantiating mapped type templates using a typedef with the
/NoTypeName/ annotation.
This work was sponsored by GOVCERT.LU.
License-Update: SIP is now licensed under the BSD-2-Clause license.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Add all relevant products from following query.
Also remove generic "protobuf" to avoid matching golang:protobuf.
.../downloads/CVE_CHECK2$ sqlite3 nvdcve_2-2.db
sqlite> select vendor, product, count(*) from products where product like '%protobuf%' group by vendor, product;
golang|protobuf|1
google|google-protobuf|5
google|protobuf|2
google|protobuf-cpp|4
google|protobuf-java|15
google|protobuf-javalite|12
google|protobuf-kotlin|6
google|protobuf-kotlin-lite|4
google|protobuf-python|4
protobuf|protobuf|1
protobuf-c_project|protobuf-c|2
protobufjs_project|protobufjs|3
rust-protobuf_project|rust-protobuf|2
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
We should not use additional vendored jsoncpp unless required due to
compatibility issues.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This commit reverts
https://git.openembedded.org/meta-openembedded/commit/?id=5af1cd3ada4cc86864810a12c82771ba1afdce64
That commit prefers sdk development instead of release images.
The additional libraries are not needed for protobuf at runtime.
Also any component depending on protobuf will have whole abseil-cpp in
sysroots so linking all the libraries is possible and any additional
ones will be installed.
What may not work is if protobuf is installed in sdk image and
application using it wants to use an additional abseil-cpp library.
However that is easily fixable by adding also abseil-cpp to sdk.
Removing hard-dependency in release images is much harder.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>