[xXPBSXx]

What

Run the liveness check before the signal 0 shortcut in pthread_kill, so signal 0 on a thread that has ended, been joined, or never existed answers ESRCH (3). The calling thread still answers 0. GuestSignals gets the cases: a live thread -> 0; once it has returned -> ESRCH for signals 0 and 15; after scePthreadJoin -> ESRCH; a bogus handle with signal 0 -> ESRCH. The debt sentence about signal 0 is removed.

Why

FreeBSD 11.4 _pthread_kill (thr_kill.c) with _thr_find_thread(..., include_dead = 0) in thr_list.c, which returns ESRCH for a dead thread.

Tested

Windows 10 Pro 19045, WinLibs GCC 15.2.0, Release. guest_signals: without the change 3/3 fail (exit 0xc0000409), with it 3/3 pass. Family guest_signals, guest_sigtimedwait, guest_kill, guest_signal_sets, guest_sigaltstack: 5/5 pass. check_conventions --base upstream/main passes. Not run: Linux, macOS, a title or console run. Console behaviour is not measured; the claim rests on FreeBSD's source only.

Fixes #8790

Overlap

#8497 also edits Signals.cpp (a different function); merge-tree against it is clean. #3986 (keep the signal mask per thread) edits Signals.cpp and GuestSignals.cpp too and conflicts with this change; whichever lands second needs a small rebase.

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 under Silent stubs in their library's technical debt file
• 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
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes pthread_kill(thread, 0) answers 0 for a thread that has ended #8790
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

AI-assisted (Claude Code).

[MarcZX10]

Reviewed exact head 79569a7a3d78b654f443b4933305f8d9a1c84def. The order now checks non-self handles with GuestThreadIsLive before the signal-0 success path, while preserving signal 0 for self. The regression covers a live thread, an ended-but-not-yet-joined thread, the joined handle, and a fabricated pointer; GuestThreadIsLive checks registry identity under lock without dereferencing an unknown handle. Issue #8790 records the Windows behavior and compares it with FreeBSD 11.4. Conventions and diff checks pass. The author reports Windows focused tests passing; hosted Linux/Windows build and runtime jobs are skipped by path policy, so this review does not claim cross-platform execution. #3986 touches the same signal/test files but implements per-thread masks; its merge/rebase interaction is separate from this behavior.

[MarcZX10]

I reviewed the exact current head 79569a7a3d78b654f443b4933305f8d9a1c84def. The liveness-before-zero ordering, self-thread special case, fake handle coverage, and ended/joined thread cases all match the reported FreeBSD behavior. Windows focused tests are reported passing; platform build jobs are skipped by path policy, so I do not infer cross-platform runtime coverage. No additional code issue found.