jeanPerier wrote: HI @mizvekov, after your patch, some llvm build broke because a piece of code in flang does not compile anymore with tip of the tree clang:
See: https://lab.llvm.org/buildbot/#/builders/4/builds/1670 Here is a reproducer: ``` #include<complex> std::complex<float> test(std::complex<float> z1) { using F0 = std::complex<float> (*)(const std::complex<float> &, const std::complex<float> &); F0 ptr{std::pow}; return ptr(z1, z1); } ``` The new error is: ``` repro.cpp:5:11: error: address of overloaded function 'pow' is ambiguous 5 | F0 ptr{std::pow}; | ^ /usr/lib/gcc/x86_64-linux-gnu/11/../../../../include/c++/11/complex:1062:5: note: candidate function [with _Tp = float] 1062 | pow(const complex<_Tp>& __x, const complex<_Tp>& __y) | ^ /usr/lib/gcc/x86_64-linux-gnu/11/../../../../include/c++/11/complex:1899:5: note: candidate function [with _Tp = float, _Up = float] 1899 | pow(const std::complex<_Tp>& __x, const std::complex<_Up>& __y) ``` Other g++ (and previously clang) compile without warning here. Is this new error justified? The piece of code with this pattern in flang is here, and is building a table of host math functions to use when folding: https://github.com/llvm/llvm-project/blob/a919588df4f108cef5829363a9ec6a1968dbb03a/flang/lib/Evaluate/intrinsics-library.cpp#L278 I have a workaround for it (#106718) but I do not understand if raising such C++ error is intentional with your patch. https://github.com/llvm/llvm-project/pull/100692 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
