siginterrupt is a libSceLibcInternal export (the public SDK's stub table has it there, not in the
kernel tables) and FreeBSD 11.4's lib/libc/gen/siginterrupt.c reads the signal's disposition,
sets or clears the signal's bit in _sigintr and SA_RESTART, and installs the disposition again.
libc cannot call libkernel, so, as the pthread hooks do, libc keeps a registered sigaction and
the address of the _sigintr set, and libkernel registers both from a static initializer when it
loads. A call before that throws; the set is written without libkernel's registration lock, as
FreeBSD's libc does. Builds on #8866, which adds the set.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The public SDK's libkernel, libkernel_sys and libkernel_web stub tables declare _sigintr, the
sigset_t that FreeBSD's libc shares between signal and siginterrupt, and no built library
provided it (#8846). libkernel now exports the 16-byte object, zero at start, and signal records
a handler with SA_RESTART unless its signal is in the set, as lib/libc/gen/signal.c does.
siginterrupt, which fills the set, is a libSceLibcInternal export; libc does not link libkernel
and cannot reach the dispositions, so it stays out and a debt entry for signal says so.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
When the native open failed with an error other than ENOENT, OpenNative
probed whether the path is a directory with std::filesystem::is_directory.
With the MinGW runtime that probe can set errno to ENOENT, and OpenNative
then reported the probe's ENOENT instead of the open error. On Windows 11
22621 a symbolic link loop failed with ELOOP from the native open and
sceKernelOpen returned 0x80020002, so guest_filesystem aborted at the
ELOOP check added by #8640.
Keep the open error and restore it when the path is not a directory.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The test throws exceptions but did not link libc, so with
-fno-asynchronous-unwind-tables it failed to link (no __gxx_personality_v0)
on the supported WinLibs toolchain; the Windows CI job does not run on pull
requests. The valid case now maps 2 MiB at a 2-MiB boundary, so the mapping is
one resident run and both Reset calls take the exact-run path, and the test
requires that run.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
pthread_kill(thread, 0) returned 0 for every handle other than the caller's before it asked whether the thread was still alive. FreeBSD's _pthread_kill looks the thread up with _thr_find_thread before it checks for signal 0, so the call answers ESRCH for a thread that has ended or was never a thread. The liveness check now runs before the signal 0 shortcut; the calling thread still answers 0.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
The crash and abort reports printed the last 64 traced calls of the whole
process, so a thread that loops on a traced import (a mutex lock/unlock
pair, a profiler marker) pushes the calls of the faulting thread out of the
report. The ring already keeps the last 65536 calls with their thread.
CopyThreadCallTrace_nid_no_patch returns the most recent calls of one
thread from that ring, oldest first, and both reports (Windows crash and
abort, Linux abort) list up to 32 of them for the reporting thread after
the process-wide list.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
