2026-09-01 - PPSA28061's wall diagnosed: no guest thread pointer (D432)
(diagnosis, no code) Turned the new fault-report bytes on PPSA28061,
the furthest title (56 imports). It faults at image+0x43c4 =
mov rax, fs:[0], read of 0x0 - the fs (thread-self) base is
zero. The install mechanism exists and is tested
(abi::thread_pointer::install, D061) but was never wired into the run
path because no title had declared a thread-local until this one. So the
next subsystem is clear: allocate a per-thread TLS block (.tdata copied,
self-pointer at [TP]), install+read-back the fs base, on the main thread
at entry and on every spawn. Not started - a wrong fs base destabilises
every run, so it wants a clean start with tests, not the end of a long
session. The _sceUltUlthreadCreate (libSceUlt user-level
threads) calls in the run are a separate subsystem behind it, not this
fault.