ExitThread went back to the thread start with longjmp on a buffer from a non-unwinding setjmp, whose frame field is 0.
When Control Flow Guard is on for the process, ucrtbase's longjmp checks the buffer first and ends the process with
fast fail 13 (0xc0000409) on a frame of 0: offset 0xb373c of ucrtbase 10.0.26100.9711, the offset in the event logs of
#222, #176 and #75. Exploit protection can force CFG on any program, and the first guest thread to exit, right after the
game's first six-axis read, killed the process with no line in the log; the abort and invalid-parameter handlers never
run for a fast fail.
plat_jmp_save and plat_jmp_go keep and restore the Win64 callee-saved registers, MXCSR and the x87 control word
themselves and validate nothing; Linux keeps setjmp and longjmp. Started with CFG forced on, the build before this
change ends with 0xc0000409 at the same log line and in --probe-kernel; this one reaches gameplay and passes the probe.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Shared runtime code now goes through runtime/platform/platform.h, implemented by platform/win32.c and a new
platform/linux.c: futex locks, a userfaultfd and PAGEMAP_SCAN write watch, signal-based async preemption, our own
fiber switch, and a SIGSEGV handler for the thaw and crash hooks. The window, input, menu and software keyboard get
SDL3 back ends beside the Win32 ones, with the swapchain (gfx_present.c) and input merging (input.c) shared. SHA-256
replaces BCrypt, ELF builds get their own override sections, and the build tools gain Linux branches.
On Linux the selftest and the gfx-abi, astc, nvn and window probes pass. probe-vkmem still fails: imported guest
memory cannot be write-watched on Linux, which the shadow buffers in vk_memory.c will address.
The FP helpers compared c->fpcr with a cached copy before every operation to resync MXCSR lazily. MXCSR now
follows FPCR.FZ at once: MSR FPCR calls rt_fpcr_sync, and so do rt_call (thread, fiber and host entry into guest
code), fiber resume (FIBER_FLAG_FLOAT_SWITCH fibers keep their own MXCSR) and the differential test harness per
case. op and vop only test FPCR.RMode for the slow path, and their redo check is one ordered-not-equal compare of
|r| with the smallest normal, the same NaN-or-smallest-normal set as before.
Hot guest code shrinks by about a quarter (132 functions holding half the guest cycles, 1.4.3); fight frame time
moves 1-3%, inside the run-to-run spread. lift, simd, func, atomic and flags_check pass on 1.4.3; simd.py fails
without the harness sync.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
