Unable to compile LLVM libc due to warnings treated as errors

[shreepo]

I'm trying to compile the LLVM libc project on Ubuntu, but I keep running into an issue where warnings are being treated as errors, specifically related to type conversion. Here is the relevant error message:

In file included from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/CPP/stringstream.h:16,
                 from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/StringUtil/error_to_string.cpp:14:
/home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/integer_to_string.h:160:58: error: conversion from ‘long unsigned int’ to ‘unsigned char’ may change value [-Werror=conversion]
  160 | template <size_t radix> using Custom = details::Fmt<radix>;
      |                                                          ^
...
cc1plus: all warnings being treated as errors

I have already tried setting the following CMake flags:

cmake -G "Unix Makefiles" -DLLVM_ENABLE_PROJECTS="libc" -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-Wno-error" -DCMAKE_C_FLAGS="-Wno-error" -DCMAKE_INSTALL_PREFIX=$HOME/local/llvm/libc ../llvm-project-18.1.8.src/llvm

And I also tried:

make CFLAGS="-Wno-error=format-truncation"

And I also tried:

env CXXFLAGS="-Wno-ignored-qualifiers -Wno-stringop-truncation -Wno-cast-function-type" make

And I also add set(CMAKE_COMPILE_WARNING_AS_ERROR OFF) at top of the llvm-project-18.1.8.src/libc/CMakeLists.txt

But the problem persists.
How can I compile the project successfully without treating these warnings as errors?
Thanks in advance for any help!

[lntue]

Instead of -DCMAKE_C_FLAGS="-Wno-error", you can add -DLIBC_WNO_ERROR=ON to your cmake command.

[shreepo]

Instead of -DCMAKE_C_FLAGS="-Wno-error", you can add -DLIBC_WNO_ERROR=ON to your cmake command.

Thank you, it worked, but another issue came up. When I configure cmake with the command：
cmake -G "Unix Makefiles" -DLLVM_ENABLE_PROJECTS="libc" -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-Wno-error -std=c++20" -DCMAKE_C_FLAGS="-std=c++20 -Wno-error" -DLIBC_WNO_ERROR=ON -DCMAKE_INSTALL_PREFIX=$HOME/local/llvm/libc -DLLVM_ENABLE_WERROR=OFF ../llvm-project-18.1.8.src/llvm
（Add -std=c++20 flag)， it gives an error:
CMake Error at /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/cmake/modules/LLVMLibCCheckCpuFeatures.cmake:72 (message): Unsupported CPU features: Call Stack (most recent call first): /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/CMakeLists.txt:300 (include)
I checked with lscpu and found that my computer does not support AVX512.it does not seem to avoid using this feature based on my CPU. Is it necessary to support AVX512?

[lntue]

is it ok for you to use ninja build instead? I don't think we have any CI bot to maintain builds with make.

If that's ok, then the simplest cmake command can be used to configure LLVM libc is:

$ cmake -GNinja -DLLVM_ENABLE_PROJECTS=libc -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=<clang/gcc> -DCMAKE_CXX_COMPILER=<clang++/g++> -DCMAKE_INSTALL_PREFIX=<...>
$ ninja libc
$ ninja install-libc

[shreepo]

is it ok for you to use ninja build instead? I don't think we have any CI bot to maintain builds with make.

If that's ok, then the simplest cmake command can be used to configure LLVM libc is:

$ cmake -GNinja -DLLVM_ENABLE_PROJECTS=libc -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=<clang/gcc> -DCMAKE_CXX_COMPILER=<clang++/g++> -DCMAKE_INSTALL_PREFIX=<...>
$ ninja libc
$ ninja install-libc

Thank you for your response. I used Ninja and configured with the following CMake command:
cmake -GNinja -DLLVM_ENABLE_PROJECTS=libc -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=~/local/gcc-14.1.0/bin/gcc -DCMAKE_CXX_COMPILER=~/local/gcc-14.1.0/bin/g++ -DCMAKE_INSTALL_PREFIX=$HOME/local/llvm/libc ../llvm-project-18.1.8.src/llvm
However, I still encounter errors during compilation when executing ninja libc. Could the issue be related to MPFR? My MPFR library is installed in a user-specific directory.
Here is the error:
[152/336] Building CXX object projects/libc/src/math/generic/CMakeFiles/libc.src.math.generic.log1pf.dir/log1pf.cpp.o FAILED: projects/libc/src/math/generic/CMakeFiles/libc.src.math.generic.log1pf.dir/log1pf.cpp.o /home/chongchuanlong/local/gcc-14.1.0/bin/g++ -DLIBC_NAMESPACE=__llvm_libc_18_1_8_ -I/home/chongchuanlong/src/LLVM/llvm-project-18.1.8.build/projects/libc/src/math/generic -I/home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/math/generic -I/home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc -isystem /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.build/projects/libc/include -fPIC -fno-semantic-interposition -fvisibility-inlines-hidden -Werror=date-time -fno-lifetime-dse -Wall -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wno-missing-field-initializers -pedantic -Wno-long-long -Wimplicit-fallthrough -Wno-maybe-uninitialized -Wno-nonnull -Wno-class-memaccess -Wno-redundant-move -Wno-pessimizing-move -Wno-noexcept-type -Wdelete-non-virtual-dtor -Wsuggest-override -Wno-comment -Wno-misleading-indentation -Wctad-maybe-unsupported -fdiagnostics-color -ffunction-sections -fdata-sections -O3 -DNDEBUG -std=c++17 -fpie -fno-builtin -fno-exceptions -fno-lax-vector-conversions -fno-unwind-tables -fno-asynchronous-unwind-tables -fno-rtti -ftrivial-auto-var-init=pattern -Wall -Wextra -Werror -Wconversion -Wno-sign-conversion -Wimplicit-fallthrough -Wwrite-strings -Wextra-semi -mavx2 -mfma -O3 -DLIBC_COPT_PUBLIC_PACKAGING -MD -MT projects/libc/src/math/generic/CMakeFiles/libc.src.math.generic.log1pf.dir/log1pf.cpp.o -MF projects/libc/src/math/generic/CMakeFiles/libc.src.math.generic.log1pf.dir/log1pf.cpp.o.d -o projects/libc/src/math/generic/CMakeFiles/libc.src.math.generic.log1pf.dir/log1pf.cpp.o -c /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/math/generic/log1pf.cpp In file included from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/integer_utils.h:15, from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/UInt.h:17, from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/UInt128.h:12, from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/FPUtil/FPBits.h:14, from /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/math/generic/log1pf.cpp:13: /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h: In function ‘__llvm_libc_18_1_8_::cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), __llvm_libc_18_1_8_::SumCarry<T> > __llvm_libc_18_1_8_::add_with_carry(T, T, T) [with T = unsigned char; cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), SumCarry<T> > = SumCarry<unsigned char>]’: /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h:77:16: error: ‘__builtin_addcb’ was not declared in this scope; did you mean ‘__builtin_addcl’? 77 | result.sum = __builtin_addcb(a, b, carry_in, &result.carry); | ^~~~~~~~~~~~~~~ | __builtin_addcl /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h: In function ‘__llvm_libc_18_1_8_::cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), __llvm_libc_18_1_8_::SumCarry<T> > __llvm_libc_18_1_8_::add_with_carry(T, T, T) [with T = short unsigned int; cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), SumCarry<T> > = SumCarry<short unsigned int>]’: /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h:86:16: error: ‘__builtin_addcs’ was not declared in this scope; did you mean ‘__builtin_addcl’? 86 | result.sum = __builtin_addcs(a, b, carry_in, &result.carry); | ^~~~~~~~~~~~~~~ | __builtin_addcl /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h: In function ‘__llvm_libc_18_1_8_::cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), __llvm_libc_18_1_8_::DiffBorrow<T> > __llvm_libc_18_1_8_::sub_with_borrow(T, T, T) [with T = unsigned char; cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), DiffBorrow<T> > = DiffBorrow<unsigned char>]’: /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h:153:17: error: ‘__builtin_subcb’ was not declared in this scope; did you mean ‘__builtin_subcl’? 153 | result.diff = __builtin_subcb(a, b, borrow_in, &result.borrow); | ^~~~~~~~~~~~~~~ | __builtin_subcl /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h: In function ‘__llvm_libc_18_1_8_::cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), __llvm_libc_18_1_8_::DiffBorrow<T> > __llvm_libc_18_1_8_::sub_with_borrow(T, T, T) [with T = short unsigned int; cpp::enable_if_t<((bool)(is_integral_v<T> && is_unsigned_v<T>)), DiffBorrow<T> > = DiffBorrow<short unsigned int>]’: /home/chongchuanlong/src/LLVM/llvm-project-18.1.8.src/libc/src/__support/math_extras.h:162:17: error: ‘__builtin_subcs’ was not declared in this scope; did you mean ‘__builtin_subcl’? 162 | result.diff = __builtin_subcs(a, b, borrow_in, &result.borrow); | ^~~~~~~~~~~~~~~ | __builtin_subcl ninja: build stopped: subcommand failed.

[lntue]

Apparently we still have some issues with building libc using gcc/g++. Is it ok for you to use clang/clang++ instead?

Otherwise, you might need to use runtime build instead, which, under the hood, build clang and use the newly built clang to build libc, with a single cmake / ninja invocation.

[nickdesaulniers]

@shreepo what does cc -v 2>&1 | tail -n 1 print for you?

error: ‘__builtin_subcb’ was not declared in this scope; did you mean ‘__builtin_subcl’?

The use of these builtins are guarded by __has_builtin, so it would seem that compiler may have a bug where it's claiming to support a given builtin, then...not.

[nickdesaulniers]

Closing since no update. Please reopen with more info provided. Thank you for taking the time to report the issue.