[nicolasabosch]

What

scePthreadCancel and pthread_cancel threw, scePthreadTestcancel did nothing and the cleanup handler exports did not exist (recorded in TechnicalDebt as "Thread cancellation is not modelled"). This implements deferred cancellation after FreeBSD libthr (thr_cancel.c, thr_clean.c, thr_exit.c).

• scePthreadCancel / pthread_cancel record the request on the thread and wake it when it is blocked in a cancellable wait. A null thread returns ESRCH.
• A request is acted on at scePthreadTestcancel, the condition variable waits, the scePthreadSem and sem waits, scePthreadJoin and the sleeps, and under their POSIX names. A condition wait takes its mutex back before the thread exits, and a cancelled join leaves the target joinable.
• The thread exits with PTHREAD_CANCELED (1) through scePthreadExit, which now runs the handlers of the new __pthread_cleanup_push_imp and __pthread_cleanup_pop_imp exports, newest first. The handler node uses the guest's _pthread_cleanup_info like libthr's struct pthread_cleanup.
• scePthreadSetcancelstate and scePthreadSetcanceltype return EINVAL for unknown values and, like libthr, test for a pending request when the type is asynchronous. pthread_setcanceltype is added.
• pthread_cancel moves from Unimplemented.cpp to Pthread/Posix/Thread.cpp with its real signature.

Limits, recorded in TechnicalDebt: a sleep is not interrupted (the thread exits when it ends, so the sceKernelUsleep path keeps its timing), file and socket calls are not cancellation points, and the asynchronous type acts like the deferred one except at the two set calls. The constants and the node layout are FreeBSD's, not measured on the console; the NIDs of the two cleanup exports come from their names and I have not seen a title import them.

#760 adds three lines to Pthread.hpp and Thread.cpp in other places; this does not depend on it.

Tested

Windows 11, WinLibs GCC 15.2.0: full build, libs target and ctest (387 passed; agc_driver_sdwa_float_selectors, shader_clock, vop1_float_unary, vop3_mad_i64 and vopc_compare excluded, they fail on this machine's Intel Arc GPU without this change). New guest_pthread_cancel test cancels threads blocked in a condition wait (while the canceller holds the mutex), scePthreadSemWait, sem_wait, scePthreadJoin and a sceKernelUsleep loop, checks the disabled state, the asynchronous type, the handler order and the error codes; it passed 40 runs in a row, and 15 with APS5_COARSE_TIMED_WAITS=1. Not tested on Linux and not tested with a title.

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
• Generic behaviour, not specific to one title
• Depends on: none
• AI-assisted: yes

[github-actions]

Progress report (948a526 → ef9d0dd)

📈 System libraries: 84.51% (+0.07%, +5 implemented functions)

✅ 5 implemented

Library	Name
libkernel	__pthread_cleanup_pop_imp_nid_postfix
libkernel	__pthread_cleanup_push_imp_nid_postfix
libkernel	pthread_cancel_nid_postfix
libkernel	pthread_setcanceltype_nid_postfix
libkernel	scePthreadCancel

[tv7]

Data point: PPSA03306 (Silt, Unity) imports __pthread_cleanup_push_imp (1xvtUVx1-Sg). On main (cea1ef13), startup stops at FAIL: unresolved ELF import 1xvtUVx1-Sg from libkernel.prx. With this PR merged locally, the import resolves and startup continues to the next missing import (hypotf, which #1196 adds). Windows 11, GCC 15.2.0, RTX 3070. I only checked startup, not whether the cleanup handlers run.

[oneandonlydean]

One thing before this can go in: please drop the // Canonical lib is libScePosix ... comment at Pthread/Posix/Thread.cpp:87-89; code comments are only allowed as technical-debt markers.

[oneandonlydean]

Conflicts with main in Pthread/src/Cond.cpp (the recursive-mutex cond wait change, 7406c97) and docs/dev/TechnicalDebt.md; please rebase.

[adriadam10]

Rebased onto current main as @oneandonlydean asked (conflicts in Pthread/src/Cond.cpp, Pthread/src/Thread.cpp and docs/dev/TechnicalDebt.md); the pthread_cancel comment was already dropped.

• Cond.cpp: the recursive-mutex wait keeps main's unlock/relock of the extra levels around ThreadCancel::Wait.
• Thread.cpp: the cancel state, cancel type, exiting flag and cleanup handler stack now live in GuestThreadState (main moved the thread state to WindowsThreadLocal) instead of separate thread_locals.
• TechnicalDebt.md: the "Thread cancellation" line replaces the old one; the neighbouring lines are main's.

Rebuilt and ran the libkernel/libs tests on Linux: all green, guest_pthread_cancel passes 30 runs in a row.

[adriadam10]

Rebased again onto current main (@oneandonlydean asked for a rebase): the only conflict was Unimplemented.cpp, where the old pthread_cancel stub and its comment are now removed since the real export lives in Pthread/Posix/Thread.cpp.

Rebuilt and ran the libkernel/libs tests on Linux: all green.

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
#2478	docs/dev/TechnicalDebt.md L325