The original reproducibility fix was problematic for a couple reasons:
- disabling both lto and parallel builds had an unfortunate effect of nearly
doubling rust-native and rust build times (which are slow to begin with).
Disabling lto hurts runtime performance too.
- both of these things were done for the *entire build*, while the only
problematic item is the librustdoc crate.
- lto=off option in config.toml has an effect only on building rustc
(the compiler itself), and doesn't help with rustdoc reproducibility.
Actual fix is the codegen-units setting, which indirectly disables
lto via giving llvm only one unit to work with at a time.
After some digging, here's a more targeted fix for the problem.
Why librustdoc is non-reproducible, but not anything else
remains a mystery, hidden deep in rust-llvm's lto optimization code.
(From OE-Core rev: a1977407a88a2004c3a6d2dba1d5bfe1aa1664b2)
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Fix markdown style issues using markdownlint tool [1]:
$ cat ./meta/recipes-devtools/rust/README-rust.md | mdl
(stdin):1: MD002 First header should be a top level header
(stdin):25: MD006 Consider starting bulleted lists at the beginning of the line
(stdin):3: MD009 Trailing spaces
(stdin):11: MD009 Trailing spaces
(stdin):26: MD009 Trailing spaces
(stdin):39: MD012 Multiple consecutive blank lines
(stdin):32: MD032 Lists should be surrounded by blank lines
(stdin):36: MD032 Lists should be surrounded by blank lines
[1] https://github.com/markdownlint/markdownlint
(From OE-Core rev: 62bade83dd6e021f11e9599a2699a1b202324f0f)
Signed-off-by: Igor Opaniuk <igor.opaniuk@foundries.io>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The valgrind test runner, vg_regtest, changed in the 3.24 update and
there are some additional valgrind ptest regressions happening with glibc-2.41.
Disable the ptests for now. Some of the remaining patches can likely
be dropped but I'd rather carry them so that re-enabling the ptests
is simpler. Also the patch series have some dependencies and unwinding
that would block the glibc-2.41 unnecessarily.
This builds for glibc, musl for qemux86-64, runs:
-> valgrind [--leak-check=full] /bin/ls /tmp
(From OE-Core rev: 8fc31fb235af12530475ec5055f6c00ccfafadfa)
Signed-off-by: Randy MacLeod <Randy.MacLeod@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
GDB 16.2 brings the following fixes and enhancements over GDB 16.1:
* PR build/32578 (cannot build GDB 16.1 out of tree when calling the configure
script with a relative path)
* PR tui/32592 ([gdb/tui] internal error in tui-winsource.c:340:refresh_window)
* PR remote/32593 (Incompatibilities between GDB's and LLDB's 'x' packet
implementation)
* PR build/32610 (Missing #include file in darwin_nat.c)
(From OE-Core rev: 11362b3ae521a0fba85c034de5ba8d4271574be3)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* Drop the CVE-2024-53589 patch, its included in 2.44
* Package newly rename gp-* to gprofng-* binaries and remove old gp-*
symlinks
* Gold linker is now deprecated and will be removed in future release
its already deleted in tarball releases
* Support for the Nios II target has been removed.
* Assembler:
- Support for new architecture extensions for AArch64, Risc-V and
x86.
* Linker:
The default maximum page size was changed from 16KiB to 64KiB for
LoongArch.
This now supports mixed LTO and non-LTO object files in
relocatable output.
The ELF forms of the linker support a --image-base=<ADDR> option
for compatibility with LLD.
The --build-id= option now accepts an argument of "xx" which
enables the use of the xxhash library. This produces a 128-bit
hash and is 2-4x faster than md5 or sha1.
The ELF linker option --package-metadata supports percent-encoded
and %[string] encoded JSON payloads.
* Disassembler:
The RISC-V disassembler now supports -M,max option like QEMU to
dump instruction without checking architecture support as usual.
* GprofNG:
Support added for hardware event counters for Neoverse-N1,
Ampere-1, and Appliedmicro processors.
Detailed release notes [1]
[1] https://lists.gnu.org/archive/html/info-gnu/2025-02/msg00001.html
(From OE-Core rev: df3c43e69542939a4bec3893f1e927edf2ad7179)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The unfs3 license matches the SPDX template for BSD-3-Clause.
There is no reason why this should be referenced as unfs3,
which is not a registered SPDX identifier.
License-Update: use SPDX identifier
(From OE-Core rev: e608747d6fba47d2bfc8637adb39ff3593559fca)
Signed-off-by: Bastian Germann <bage@debian.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The EDL-1.0 license matches the SPDX template for BSD-3-Clause.
There is no reason why this should be referenced as EDL-1.0, which is
not a registered SPDX identifier for exactly this reason.
License-Update: use SPDX identifier
(From OE-Core rev: 9745740bbab58365f86ef579e1ad1e86b90594a6)
Signed-off-by: Bastian Germann <bage@debian.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The XSL license was standardized by SPDX as DocBook-XML.
Copy the license file to the new name and use it in
docbook-xsl-stylesheets.
License-Update: use new DocBook-XML SPDX ID, copy text from existing XSL
(From OE-Core rev: e172e27e19fc09160cfee2a29af8e8c5c2426c9c)
Signed-off-by: Bastian Germann <bage@debian.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Recipes are much more readable with whitespace around the assignment operators.
Fix various assignments in OE-Core to show this is definitely the preferred
formatting.
(From OE-Core rev: 30ea609d3357fb3de911f2f6a5e6856c151b976a)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
YOCTO [#15061]
The rust target and linker are getting setting from the sdk environment and
so the config file is not needed. The redundant config file geneartion is removed.
(From OE-Core rev: d5f78816d2ad0f3e43ce883eef199d1683cfcbb4)
Signed-off-by: Harish Sadineni <Harish.Sadineni@windriver.com>
Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
YOCTO [#15061]
The rust sdk installs both 'rust.sh' and 'cargo.sh' for lib32 and lib64 in the same location.
This causes below error while installing the lib32 & lib64 binaries:
Error: Transaction test error:
file /usr/local/oe-sdk-hardcoded-buildpath/sysroots/x86_64-pokysdk-linux/environment-setup.d/cargo.sh
conflicts between attempted installs of rust-cross-canadian-arm-1.67.1-r0.x86_64_nativesdk and
rust-cross-canadian-aarch64-1.67.1-r0.x86_64_nativesdk
file /usr/local/oe-sdk-hardcoded-buildpath/sysroots/x86_64-pokysdk-linux/environment-setup.d/rust.sh
conflicts between attempted installs of rust-cross-canadian-arm-1.67.1-r0.x86_64_nativesdk and
rust-cross-canadian-aarch64-1.67.1-r0.x86_64_nativesdk
ERROR: Task (virtual:multilib:lib32:/media/build/poky/meta/recipes-sato/images/core-image-sato.bb:do_populate_sdk)
failed with exit code '1'
The change includes:
- Prepending '${RUST_TARGET_SYS}' to 'rust.sh' to differentiate between target systems.
- Moving the non-target-specific environment variables to 'nativesdk-cargo' and 'nativesdk-rust',
instead of being managed by the cross-canadian recipe.
(From OE-Core rev: 40eb4bfe2f100ba5301046ca25110fcc55a640bb)
Signed-off-by: Harish Sadineni <Harish.Sadineni@windriver.com>
Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The string package in python currently contains common string
operations. This package is still in full use.
However, the old summary line of the package marked it as deprecated. As
this is not the case, this commit updates the summary line to better
represent this Python package.
The new summary line was taken from Python's documentation for the
package (https://docs.python.org/3.13/library/string.html).
In the past, the string library used to contain functionality that later
was moved to the str object in Python. The assumption is that during
this change, this library was marked as deprecated here. However, as
this package contains more functionality, which is still relevant and
used by python, the package itself should not be marked as deprecated.
This commit changes the description of the package, but does not change
the name of the package. The proper name for this package should be
"string" (instead of "stringold"). However, changing the package's name
will break compatibility with anyone using the string package in their
project. As this is a big change, it is decided not to be done yet.
The change in description, as well as this commit message, should be
enough to let users understand that the string package in python is
still valid, and should not be treated as deprecated.
(From OE-Core rev: 7602a9013d6c61959ec293ee29285c90257b5fbb)
Signed-off-by: Omri Sarig <omri.sarig13@gmail.com>
Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Summary of changes:
- New custom dependency for atomic
- --cap-lints allow used for Cargo subprojects
- Cargo features are resolved globally
- Meson can run "clippy" on Rust projects
- Devenv support in external project module
- Fixed sizeof and find_library methods for Fortran compilers
- format command now accept stdin argument
- "machine" entry in target introspection data
- Add new language Linear Asm
- Control the number of child processes with an environment variable
- Support for Rust 2024
- Support TASKING VX-Toolset
- Test targets no longer built by default
- Install vcs_tag() output
(From OE-Core rev: 37d4a5b2653b94d5030b5026b1ecf979420cfb20)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When using the -fsanitize=address CXX_FLAG for a program compiled for
aarch64 / arm64
This is happing:
MemorySanitizer: CHECK failed: sanitizer_allocator_primary64.h:133 "((kSpaceBeg))
== ((address_range.Init(TotalSpaceSize, PrimaryAllocatorName, kSpaceBeg)))"
(0xe00000000000, 0xfffffffffffffff4) (tid=51745)
With -DSANITIZER_CAN_USE_ALLOCATOR64=0 this is not happening and
potenial bugs are detected.
ARM32 does not require this patch.
More info about the issue in this thread:
https://github.com/llvm/llvm-project/issues/65144
(From OE-Core rev: 12442b9b6df06317174066854935b1d6a4f1865d)
Signed-off-by: Thomas Roos <throos@amazon.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Ross Burton <ross.burton@arm.com>
zipfs is a new facility in tcl 9.x where various data files are bundled
into a zip archive, rather being separately installed.
Then that zip is embedded into libtcl.so from Makefile, thusly:
cat ${TCL_ZIP_FILE} >> ${LIB_FILE}
This is a major case of face meeting palm: any binary object
processing on the resulting .so file discards the extra data
at the end, and that's exactly what happens in do_package(),
resulting in a tcl installation without any language libraries.
This is not caught by ptest because it runs against a private
copy of the source tree.
Additionally, it helps to have data files on target systems
as files that can be viewed and edited.
(From OE-Core rev: 05e31be56498123b177f363c700c96b20958585c)
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Ross Burton <ross.burton@arm.com>
We have a patch to allow us to 'poison' system include directories,
which are warnings by default but we make them fatal in cross builds.
However, in the 13.1 upgrade[1] the patch to make the warnings fatal was
dropped in the compiler invocation, so it only took effect for pure
preprocessor calls. This was not noticed at the time as the test case
was flawed, but this has now been fixed.
Add back the fatal poisoning, and restructure the patch slightly so it
is less invasive.
[1] oe-core bea46612fd9106cc5b46eb1d81623b6492563c13
[RP: Tweak to fix gcc/gcc-cross-canadian failure]
(From OE-Core rev: 56f21a02c009cb74072ee79467a5bcab3c4643a5)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With gcc posioning fixed, this recipe showed errors, using an incorrect include
path looking at the host system. If pkgconfig is present, the correct include
paths are used. Therefore add the missing dependency.
(From OE-Core rev: 6cf0aaa3af276694709369b6007f629862e21559)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently, providers are set on a global config basis. This change allows
for a select set of providers to be configured using BB_RECIPE_VIRTUAL_PROVIDERS
on a per recipe basis. This would allow for the selection of virtual/cross-cc
as gcc or clang for example.
The PROVIDERS are removed from the recipes so that if a version of the
dependency accidentally slips through, the build will fail and the user
can correct the issue.
(From OE-Core rev: 6eeab1a5d7f23917b94c130e417d59afb757b546)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
