spin build -j 12 builds just fine for me locally on ARM Mac on latest main with Python 3.12. Can you provide a bit more information on your reproducible build failure on Mac? Your ticket/issue is a bit hard for me to parse--I do see a bleeding edge 3.15.0a0 version of CPython there--is that the key to reproducing?

Perhaps the python version is relevant, but with 3.12 I get a similar error.

06:14:33-0600[numpy@main]38208$ python vendored-meson/meson/meson.py compile -C build3.12 INFO: autodetecting backend as ninja INFO: calculating backend command to run: /Users/wrp/x86_64/Darwin/bin/ninja -C /Users/wrp/github/numpy/numpy/build3.12 ninja: Entering directory /Users/wrp/github/numpy/numpy/build3.12'
[223/523] Compiling C object numpy/_core/libloops_exponent_log.dispatch.h_AVX2.a.p/meson-generated_loops_exponent_log.dispatch.c.o
../numpy/_core/src/umath/loops_exponent_log.dispatch.c.src:51:1: warning: unused function 'fma_get_full_load_mask_pd' [-Wunused-function]
fma_get_full_load_mask_pd(void)
^
../numpy/_core/src/umath/loops_exponent_log.dispatch.c.src:66:1: warning: unused function 'fma_get_partial_load_mask_pd' [-Wunused-function]
fma_get_partial_load_mask_pd(const npy_int num_elem, const npy_int num_lanes)
^
../numpy/_core/src/umath/loops_exponent_log.dispatch.c.src:83:1: warning: unused function 'fma_masked_gather_pd' [-Wunused-function]
fma_masked_gather_pd(__m256d src,
^
../numpy/_core/src/umath/loops_exponent_log.dispatch.c.src:98:1: warning: unused function 'fma_masked_load_pd' [-Wunused-function]
fma_masked_load_pd(__m256i mask, npy_double* addr)
^
../numpy/_core/src/umath/loops_exponent_log.dispatch.c.src:110:1: warning: unused function 'fma_set_masked_lanes_pd' [-Wunused-function]
fma_set_masked_lanes_pd(__m256d x, __m256d val, __m256d mask)
^
5 warnings generated.
[268/523] Compiling C++ object numpy/_core/libunique_hash.a.p/src_multiarray_unique.cpp.o
FAILED: [code=1] numpy/_core/libunique_hash.a.p/src_multiarray_unique.cpp.o
c++ -Inumpy/_core/libunique_hash.a.p -Inumpy/_core -I../numpy/_core -Inumpy/_core/include -I../numpy/_core/include -I../numpy/_core/src/common -I/Users/wrp/.local/share/uv/python/cpython-3.12.9-macos-x86_64-none/lib/pkgconfig/../../include/python3.12 -I/Users/wrp/github/numpy/numpy/build3.12/meson_cpu -fdiagnostics-color=always -Wall -Winvalid-pch -std=c++17 -O2 -g -ftrapping-math -DNPY_HAVE_CLANG_FPSTRICT -msse -msse2 -msse3 -DNPY_HAVE_SSE2 -DNPY_HAVE_SSE -DNPY_HAVE_SSE3 -DNPY_INTERNAL_BUILD -DHAVE_NPY_CONFIG_H -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE=1 -D_LARGEFILE64_SOURCE=1 -fexceptions -fno-rtti -MD -MQ numpy/_core/libunique_hash.a.p/src_multiarray_unique.cpp.o -MF numpy/_core/libunique_hash.a.p/src_multiarray_unique.cpp.o.d -o numpy/_core/libunique_hash.a.p/src_multiarray_unique.cpp.o -c ../numpy/_core/src/multiarray/unique.cpp
../numpy/_core/src/multiarray/unique.cpp:243:36: error: implicit instantiation of undefined template 'std::vector<npy_unpacked_static_string>'
std::vector<npy_static_string> unpacked_strings(isize, {0, NULL});
^
/Library/Developer/CommandLineTools/SDKs/MacOSX12.1.sdk/usr/include/c++/v1/iosfwd:248:28: note: template is declared here
class _LIBCPP_TEMPLATE_VIS vector;
^
1 error generated.
[273/523] Compiling C object numpy/_core/libloops_autovec.dispatch.h_AVX2.a.p/meson-generated_loops_autovec.dispatch.c.o
ninja: build stopped: subcommand failed.
06:17:36-0600[numpy@main]38208$ python --version
Python 3.12.9`

Instead of the python version, it's more likely related to clang, which is from llvm-project@87f0227cb60147a26a1eeb4fb06e3b505e9c7261. I'll try to find time to play around and narrow down the issue.