[mugurc]

What

nid_patcher turns the _nid_postfix names of a library into NIDs for ELF and PE. This adds Mach-O, so that a library built on macOS gets its NIDs too. It is the piece of the macOS port that ninja libs needs there; the build changes that use it come in separate PRs. Nothing changes for ELF and PE.

• A Mach-O dylib keeps its exports in a dyld export trie and its imports in the chained fixups, so the patcher works on those. The trie is read, every exported C name (without Mach-O's leading underscore) is resolved to its NID, and the trie is rebuilt the way ld64 lays it out: node offsets are uleb128, so the layout is repeated until it is stable. A trie that no longer fits is appended to __LINKEDIT, which has to end the file, and a signed image is refused.
• Weak definitions that are not marked _nid_postfix or _nid_no_patch keep their names, because dyld coalesces those across images by name.
• In the chained imports, names bound to a .prx dylib that are guest facing (a _nid_postfix name or an sce prefix) are patched in place. A name that would grow, or that shares storage with another import, is an error. An image that uses bind opcodes instead of chained fixups is refused.
• --preserve-exports reads a reference Mach-O the way it reads ELF and PE references.

Tested

• New test nid_patcher_macho: builds small Mach-O dylibs byte by byte and runs nid_patcher on them. It checks the rebuilt export trie (including a weak export that must stay), the patched and untouched chained imports, a trie that grows and is appended to __LINKEDIT, --preserve-exports, and the three refusals (signed image, bind opcodes, an import NID longer than its name). It needs no macOS. I also ran it against two copies of the patcher with a deliberate break each (no import patching, weak exports patched), and both fail it.
• macOS 26.5 (M5 Max): the patcher turns all 121 libraries of the macOS port into the same bytes the port builds, checked with cmp, and the port's patched_libraries test loads them all.
• Linux and Windows: the full build and ctest, run by manual dispatch on my fork: https://github.com/mugurc/AnyPS5/actions/runs/37931715161 (Linux 508 of 508, Windows 517 of 517, nid_patcher_macho included). The relinker jobs on a pull request do not build nid_patcher, so they do not cover this change.
• No commercial title was tested; I have no PS5.

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
• AI-assisted: yes

[popcoder565]

Native Windows run, since no Windows results were posted. Setup: Windows 10 (19045), MinGW-w64 GCC 15.2.0 (WinLibs r7), Release. Head f3c51474, merged onto main d70b899 together with #1324, #2039 and #2193.

• nid_patcher_macho passes on Windows.
• Full ctest in that build: 512 of 512. That needed two local workarounds in feat(libs): implement font, libc, libkernel and misc prx stubs #2193, which doesn't build on Windows (reported there).

🤖 Addressed by Claude Code

[adriadam10]

Checked the build and ctest scoped to core/libs (11/11, nid_patcher_macho included), a clean merge onto current main, and popcoder565's native Windows run of the new test. ELF/PE paths are unchanged and the Mach-O test pins the rebuilt trie, the patched and untouched chained imports, --preserve-exports and the three refusals. Merging.