Clang交叉编译arm-none-eabi目标时无法正确识别ARM GNU Toolchain的问题求助

Let's break down the two main issues you're facing and fix them step by step:

1. Linker Error: Missing ld.lld #

Clang defaults to using LLVM's lld linker for cross-compilation, but your ARM GNU Toolchain only includes GNU's ld.bfd linker. To resolve this, explicitly tell Clang to use ld.bfd with the -fuse-ld=bfd flag.

2. Incorrect Header Search Paths #

Clang is looking for libc++ (LLVM's standard library) headers at include/c++/v1, but your ARM Toolchain provides GCC's libstdc++ headers at include/c++/15.2.1. You need to:

• Switch Clang to use libstdc++ instead of libc++ with -stdlib=libstdc++
• Fix the --gcc-toolchain parameter: it should point to the root directory of your ARM toolchain (not the sysroot subdirectory), which helps Clang locate GCC's standard libraries and tooling correctly.

Working Compilation Command #

Putting it all together, here's the corrected command:

clang --verbose \
--target=arm-none-eabi \
--gcc-toolchain=${HOME}/Desktop/arm-gnu-toolchain-15.2.rel1-x86_64-arm-none-eabi \
--sysroot=${HOME}/Desktop/arm-gnu-toolchain-15.2.rel1-x86_64-arm-none-eabi/arm-none-eabi \
-fuse-ld=bfd \
-stdlib=libstdc++ \
-o executable \
main.cpp

Why This Works #

• --gcc-toolchain: Tells Clang where to find the ARM GCC toolchain's auxiliary files (like libstdc++ headers and runtime libraries). This parameter was being ignored before because you paired it with a sysroot path but didn't switch to libstdc++.
• -stdlib=libstdc++: Forces Clang to search for GCC's standard library headers instead of libc++, so it will look for include/c++/15.2.1 instead of include/c++/v1.
• -fuse-ld=bfd: Directs Clang to use the GNU linker included in your toolchain instead of trying to invoke lld, which isn't present.

Bonus Tip #

For real-world ARM embedded projects, you'll likely need to specify a target CPU (e.g., -mcpu=cortex-m4 for a Cortex-M4 core) or floating-point ABI (e.g., -mfloat-abi=hard if your core has an FPU). Add these flags as needed for your hardware.

内容的提问来源于stack exchange，提问作者NoodleCollie