[mugurc]

What

The parts of libc that talk to the host and do not build with Darwin's headers and libSystem. Part of the macOS series announced on #1 (#1 (comment)).

• feof_unlocked, ferror_unlocked, getc_unlocked and putc_unlocked are macros on macOS, so they are called without the :: qualifier.
• CLOCK_REALTIME_COARSE and st_mtim do not exist on macOS: the write mark of General.cpp uses CLOCK_REALTIME and st_mtimespec.
• tm_zone is a char*, not a const char*.
• libSystem's __cxa_thread_atexit returns void; the wrapper returns 0.
• Darwin's stack protector reads __stack_chk_guard from libSystem, and the library's own definition under that name would be bound by every other library's host code. On macOS the guard has the name __stack_chk_guard_nid_postfix, which the NID patcher gives the NID of the original.
• APS5_EXPORT spells its symbols with the leading underscore of Mach-O.
• The fclose failure check of GuestFiopen fills /dev/full, which exists on Linux only.
• exception_runtime is registered everywhere but macOS (the follow-up I promised on feat(libc): exception runtime on macOS hosts #927). It expects a host throw to go through libc's exception runtime, which libc takes over on Linux by exporting the host runtime names; on macOS it does not export them, and the system libc++abi handles the throw.
• The formatting test names GCC's __builtin_sysv_va_list, which clang does not have; for clang it is va_list, the System V one on the hosts where it is used.

Tested

• macOS 26.5.1, Apple M5 Max, Apple clang 21, x86-64 build under Rosetta 2. main 7e2ba73 with the rest of the macOS series (ci/macos-series-2 on my fork, which now includes feat(libkernel): JIT shared memory on macOS hosts #7937) and the libSceSsl change of feat: macOS on Apple Silicon host support (Rosetta 2 and MoltenVK) #7332. The libraries do not link on macOS without the series. ninja, ninja libs and ctest -j4: 1798 tests, 1123 passed, 25 failed, the rest skipped (no Vulkan loader in the environment); guest_formatting, guest_fiopen and guest_fopen_errors pass.
• Rebased on main 7e2ba73. The conflict was getc_unlocked_nid_postfix, which fix(libc): switch update streams to reading as FreeBSD does #8226 now reads through a local native; the call stays unqualified. The GuestArena.cpp change is gone from this pull request, because feat(libc): reserve the guest arena on macOS #6124 builds the same code on macOS.
• Earlier, with this pull request alone on main: the changed libc sources and the formatting and file stream tests compiled on macOS, and they did not compile without it.
• Linux and Windows: not run for this version. The only edits outside __APPLE__ are the four unqualified calls (the functions are declared in the global namespace by glibc), the moved test registration (the test is still registered on Linux), the va_list mapping in the test for clang, and #ifdef __linux__ in place of #ifndef _WIN32 around the /dev/full check (Windows skipped it already). An earlier version ran on my fork's CI with the rest of the series: 960 tests on Linux and 966 on Windows pass (https://github.com/mugurc/AnyPS5/actions/runs/38256280627).

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed in TechnicalDebt
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source (none added)
• Generic behaviour, not specific to one title
• Depends on: none
• Unblocks: none
• AI-assisted: yes

[MarcZX10]

I opened #5394 to carry the Windows line-buffering fix from #5391. It touches putc_unlocked_nid_postfix for newline flushing; I listed #4627 as a dependency so the macOS portability change can land first, then I can rebase and resolve the shared lines. If you prefer a different merge order, please say so.

[krishalivaibhav]

Code review: the portability changes read correctly. The macros behind getc_unlocked/feof_unlocked explain the dropped ::. tm_zone needs the const_cast on Apple. __cxa_thread_atexit returns void there, so the unconditional return 0 is right. The APS5_EXPORT underscore prefix matches Mach-O symbol naming. Could you add the macOS build output and the libc ctest results to the description? MarcZX10's #5394 lists this PR as a dependency, so a review from a maintainer would unblock both. (Diff review only.)