Add the GBK (CP936) and EUC-CN profiles, their scalar conversions to
and from UTF-8, UTF-16, UTF-32 and UCS-2, and accept both profiles in
the MBCS string conversions.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Implement sceCesSbcTo{Ucs2,Utf16,Utf32}* and their inverses, and the
UTF-16 and UTF-32 MBCS string conversions and length queries, on the
existing single-byte, CP932 and CP51932 profiles.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Implement the single-character conversions between the CP932 and
CP51932 profiles and UTF-8, UTF-16, UTF-32 and UCS-2 in both
directions, and the Shift-JIS to JIS X 0208 row/cell arithmetic.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Measured on firmware 13.60: attribute 0x100 makes a flag openable by
name, and only such flags can be opened. Each open returns a new handle
on the same flag, which sceKernelCloseEventFlag releases. Closing a
creator handle and deleting through an opened one return EPERM. Two live
shared flags cannot share a name, names of 32 or more bytes return
ENAMETOOLONG, and a missing name returns ESRCH. After the creator
deletes the flag, an opened handle polls and waits with EACCES and sets,
clears and cancels with no effect.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
fdfawef324 notes that current LLVM spills SGPRs into VGPRs it allocates like any other
register, so one window register can be an ordinary register early in a
program and hold spill lanes later. ReadLaneEliminator only followed a
spilled lane across M0-indexed window stores for slots never used as
ordinary registers, so RESIDENT EVIL 4 (PPSA07411) and LEGO Horizon
Adventures (PPSA14632) kept their reloaded descriptors as run-time
values and stopped. A known lane can only come from a v_writelane chain,
so keep it across a computed-index store instead; a constant-index store
of the slot still ends the spill. This is fdfawef324's suggestion on
#6277.
That such a store misses the spill cannot be proven for these programs,
so it is checked at run time where a replaced read depends on it: a
backward liveness over each slot and lane whose known value a replaced
read used puts a WindowLaneGuard before every computed-index store at
which that lane is live, and the guard records a WindowLaneSpill fault
when the store changes that lane of the slot. The driver reads the fault
buffer back for every program with such a guard and reports the dispatch
or draw as failed, so a broken assumption is no longer silent. This also
covers the pure spill slots main already folds.
Co-authored-by: fdfawef324 <[email protected]>
Co-Authored-By: Claude <[email protected]>
The entry was twice the 1000-character limit, led with the behaviour
rather than the gaps, and named a title. It now lists what throws and
what is inferred first, and names no title.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Vulkan needs PointSize above zero. A PA_SU_POINT_SIZE of 0 without
USE_VTX_POINT_SIZE, and a PA_SU_POINT_MINMAX maximum of 0 with it, now
throw. With USE_VTX_POINT_SIZE the clamp's lower bound is at least the
smallest positive float, so an exported size of 0 or less, which a
clamp minimum of 0 lets through, is written as a positive value.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A point list routed through the vertex path was refused, and so was
PA_CL_VS_OUT_CNTL.USE_VTX_POINT_SIZE on it. OliSe Player (PPSA01153)
stops on the latter with two of its MilkDrop presets.
The vertex path now draws point lists as Vulkan points. The vertex stage
writes PointSize: the misc export's X clamped to PA_SU_POINT_MINMAX when
USE_VTX_POINT_SIZE is set, and the PA_SU_POINT_SIZE height otherwise.
The point state travels in the recompile request (version 31) and the
cache and draw keys. Non-square register sizes, inverted clamps, point
sprite texture coordinates and, without largePoints, sizes other than
one pixel throw.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Measured on firmware 13.60: create accepts attribute bits 0x1, 0x2, 0x10,
0x20 and 0x100 and rejects 0x1|0x2; wait and poll ignore mode bits outside
the match and clear fields; EPERM, ECANCELED and EACCES write 0 to the
result pattern.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
On Windows a pair end now reports POLLIN after the peer's shutdown or close.
The console also reports POLLHUP after close; the loopback connection cannot
tell a close from a shutdown.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
FreeBSD 11 kqueue_scan (sys/kern/kern_event.c) re-inserts a reported
knote without EV_CLEAR or EV_DISPATCH at the tail of kq_head, and
knote_enqueue puts a knote that fires again at the tail as well, so the
next scan starts with the events that were not reported yet.
sceKernelWaitEqueue scanned the event list in place, so with num smaller
than the number of ready events the first ones were reported on every
call and the later ones starved. Each reported event that stays
registered now moves to the end of the list, in the order reported.
FreeBSD 11's kern_sigaction (sys/kern/kern_sig.c) calls
sigqueue_delete_proc, which drops the signal from the process queue and
from every thread's queue, whenever the new handler is SIG_IGN, or
SIG_DFL on a signal whose sigprop has SA_IGNORE (SIGURG, SIGCONT,
SIGCHLD, SIGIO, SIGWINCH, SIGINFO). It does so whether or not the
signal is blocked. sigaction and signal now take the signal off the
pending set in those cases, so a later unblock or wait never sees it.
FreeBSD 11.4 getsockaddr sets sa_len to the length argument, and
tcp_usr_bind, tcp_usr_connect, in_pcbbind_setup, in_pcbconnect_setup,
in6_pcbbind and in6_pcbladdr answer EINVAL when it is not the size of
the socket family's address. udp6_bind and udp6_connect pass an
IPv4-mapped address to the IPv4 code without that check when
IPV6_V6ONLY is clear, and udp6_send always checks it. bind and connect
on AF_INET and AF_INET6 sockets, and sendto on AF_INET6 datagram
sockets, now fail a longer address with EINVAL.
WSAPoll reports POLLHUP alone for a stream socket whose peer has closed, so
poll never returned POLLIN for it and a caller waiting for POLLIN did not see
the end of the stream. sopoll_generic reports a socket that can receive no
more as readable and adds POLLHUP only once it can send no more either.
poll now adds the requested read bit and keeps POLLHUP only after the socket
shut its own sending side. A reset or refused connection keeps WSAPoll's flags.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
FreeBSD 11.4's soreceive_generic waits in sbwait (sys/kern/uipc_socket.c,
uipc_sockbuf.c), whose msleep_sbt on so_rcv.sb_timeo returns EWOULDBLOCK
(35) for every socket type when SO_RCVTIMEO expires. On Windows the host
socket is non-blocking and native_transfer enforces the timeout itself,
but it reported WSAETIMEDOUT, which only datagram receives turned into
EAGAIN; a stream receive failed with ETIMEDOUT (60). A receive that
reaches the stored timeout now reports WSAEWOULDBLOCK, so every
receive returns EAGAIN, and a host WSAETIMEDOUT still maps to ETIMEDOUT.
FreeBSD 11 sendfile checks the file descriptor first, then looks up the
socket with getsock_cap, which answers ENOTSOCK for an open descriptor
that is not a socket and EBADF for one that is not open
(sys/kern/kern_sendfile.c, sys/kern/uipc_syscalls.c).
sendfile replaced the errno that Lookup had already set with EBADF. It
now reports the errno Lookup chose, so an open non-socket gives ENOTSOCK
and a closed descriptor still gives EBADF. The sentence about this gap
is cut from the sendfile debt entry and guest_sendfile covers both cases.
Both threw for a regular file. They now follow FreeBSD 11.4's
filt_vfsread and filt_vfswrite: a read event reports the file size less
the descriptor's offset and fires while that is not 0; a write event
always fires with data 0. A wait holding such an event polls it again
every millisecond, as no host call reports every file change. A size
above 1 and other vnode types still throw.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Federico Marras