linux/poky.git/meta/recipes-devtools/gcc, branch 5.2_M3

gcc-sanitizers.inc: Workaround for aarch64

2025-01-23T11:56:47+00:00

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 Signed-off-by: Richard Purdie Signed-off-by: Ross Burton

gcc: make include poisoning fatal again in gcc/g++

2025-01-22T13:20:29+00:00

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 Signed-off-by: Richard Purdie

base: Switch virtual/cross-XXX to be under recipe specific providers

2025-01-21T23:09:34+00:00

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

classes/recipes: Switch virtual/XXX-gcc to virtual/cross-cc (and c++/binutils)

2025-01-21T23:09:33+00:00

The idea of the base class dependency is to say "yes, I need a C cross compiler" and this was never meant to be gcc specific. Looking at the codebase, whilst we code triplets into this, it does overcomplicate things as there are only ever limited, "target", "sdk" and the class extended versions like mutlilib. After much thought, we can simplify this to virtual/cross-cc and virtual/nativesdk-cross-cc. This lets us remove the "gcc" specific element as well as removing the over complicated triplet usage. At the same time, change the much less widely used "g++" variant to "c++" for similar reasons and remove the triplet from virtual/XXX-binutils too. Backwards compatibility mappings could be left but are just going to confuse things in future so we'll just require users to update. This simplification, whilst disruptive for any toolchain focused layers, will make improved toolchain selection in the future much easier. Since we no longer have overlapping variables, some code for that can just be removed. The class extension code does need to start remapping some variables but not the crosssdk target recipe names. This patch is in two pieces, this one handles the renaming with the functional changes separate in a second for easier review even if this breaks bisection. (From OE-Core rev: 4ccc3bc8266c327bcc18c9a3faf7536210dfb9f0) Signed-off-by: Richard Purdie

gcc: poison-system-directories patch updated for missing paths

2025-01-20T13:38:59+00:00

Modified logic in gcc/incpath.cc to ensure that non-existing host system paths are not deleted during cross-compilation. If the build system attempts to search a host path, gcc will now issue a warning instead of silently ignoring it. Fixes [YOCTO #15672] https://bugzilla.yoctoproject.org/show_bug.cgi?id=15672 (From OE-Core rev: 32e2f9a73a653a00e555e50f9197b605f2f70f89) Signed-off-by: Sunil Dora Signed-off-by: Richard Purdie

qemu/machine: rename QEMU_EXTRAOPTIONS for consistency

2025-01-11T18:34:15+00:00

The per-tune qemu options variable is QEMU_EXTRAOPTIONS_${TUNE_PKGARCH}, but this doesn't follow the pattern of all of the other tune-specific variables in the machine configuration which is VARIABLE:tune-[name]. Rename QEMU_EXTRAOPTIONS_${TUNE_PKGARCH} to QEMU_EXTRAOPTIONS:tune-${TUNE_PKGARCH} for consistency. Note that this will mean that BSPs need to update any assignments of this variable. (From OE-Core rev: 7f981d074442b901f7e64dbdb9db851ff31c3733) Signed-off-by: Ross Burton Signed-off-by: Richard Purdie

classes/qemu: use tune to select QEMU_EXTRAOPTIONS, not package architecture

2025-01-11T18:34:15+00:00

Using the package architecture to select the right qemu options to pass to qemu-user is incorrect, and fails for recipes that set PACKAGE_ARCH to MACHINE_ARCH (as the qemuppc workarounds suggest) because there are not typically any options set for the machine name. Solve this by using TUNE_PKGARCH instead: for the majority of recipes this is the same value, but for machine-specific recipes it remains the same instead of changing to the machine name. This means we can remove the qemuppc workarounds, as they're obsolete. Also update the gcc-testsuite recipe which uses the same pattern to use TUNE_PKGARCH, and generalise the else codepath to avoid needing to update the list of architectures. [ YOCTO #15647 ] (From OE-Core rev: 414b754a6cbb9cc354b1180efd5c3329568a2537) Signed-off-by: Ross Burton Signed-off-by: Richard Purdie

gcc: backport patch to fix data relocation to !ENDBR: stpcpy

2024-12-18T11:11:55+00:00

There is the following warning when building linux-yocto with default configuration on x86-64 with gcc-14.2: AR built-in.a AR vmlinux.a LD vmlinux.o vmlinux.o: warning: objtool: .export_symbol+0x332a0: data relocation to !ENDBR: stpcpy+0x0 This change set removes the warning. PR target/116174 [https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116174] (From OE-Core rev: 30d4f18d1e11b3336c8668dccd96b9ff35c7bc76) Signed-off-by: Bin Lan Signed-off-by: Richard Purdie

gcc: Fix c++: tweak for Wrange-loop-construct

2024-12-13T12:26:01+00:00

This commit updates the warning to use a check for "trivially constructible" instead of "trivially copyable." The original check was incorrect, as "trivially copyable" only applies to types that can be copied trivially, whereas "trivially constructible" is the correct check for types that can be trivially default-constructed. This change ensures the warning is more accurate and aligns with the proper type traits. LLVM accepted a similar fix: https://github.com/llvm/llvm-project/issues/47355 PR c++/116731 [https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116731] (From OE-Core rev: 7c9402953d1e4eed4bd4427171e96761812f7478) Signed-off-by: Marek Polacek Signed-off-by: Sunil Dora Signed-off-by: Mathieu Dubois-Briand Signed-off-by: Richard Purdie

gcc-sanitizers: Add riscv64 as compatible host

2024-12-03T11:21:34+00:00

RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57), but ASan currently supports only Sv39 on RISC-V64. For RISC-V64 Sv39, ASan uses custom allocator configurations tuned to manage large allocations efficiently. These tunings are incompatible with larger address spaces like Sv48/Sv57 due to differences in region sizes and alignment. For riscv64, Asan's tuning for Sv39 can be enabled in qemu by using the appropriate flag in the command line as shown below: runqemu nographic qemuparams="-cpu rv64,sv39=true" Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been updated to include riscv64. All necessary tests were successfully conducted on both hardware(Microchip PolarFire SoC) and the qemurisv64 environment, with ASan effectively detecting memory errors in both scenarios. (From OE-Core rev: 4b4450ff695ef73bf7a2437e142d2e0730d6a547) Signed-off-by: Deepthi Hemraj Signed-off-by: Mathieu Dubois-Briand Signed-off-by: Richard Purdie