A FLAT_SHADE input without pass-through that a pixel shader reads
through v_interp_mov_f32 P20 and not P10 stayed an explicit per-vertex
input and read vertex 2 minus vertex 0. For a flat value that is the same
on every vertex the difference is 0, and with the first provoking vertex
nothing refused it.
ps5-opengl lowers every flat input to a load of one vertex: vertex 1
(P10) by default, which #4030 reads as the provoking vertex, and vertex 2
(P20) for triangles when GL's flatshade_first is set. SuperTuxKart's
deferred point lights are drawn that way, so all twelve flat inputs of
pointlight.frag read 0, the attenuation 0 * -inf was NaN, nothing was
discarded, and the starting-line referee's light quad blacked out the
scene behind it (#8364).
A P20-only read now returns the provoking vertex's value, as P10 and P0
do. The lane was not measured on the console either, so FragmentParameter
records it and a pipeline using the last provoking vertex throws. An input
read through both P10 and P20 keeps its per-vertex differences.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
PPSA03396 calls sceAgcAcbMemSemaphore and sceAgcDcbMemSemaphore as
(cb, address, waitForMailbox, signalType, operation). Both read the
address from the fourth argument instead, so the call aborted with
"null or misaligned address 0x1".
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sceAudioOut2SetSystemDebugState accepts only the loudness debug state (id 1, 8-byte param); other ids and sizes return the invalid param error and a null pointer returns the invalid pointer error. The one debug state the console also accepts, id 0x20 with a 4-byte param, throws.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
An unsupported speaker count returns 0x804A8451, a unit angle other than 360 or 2*pi with angles returns 0x804A8450 and a null work returns 0x804A8010, leaving the work untouched, instead of throwing.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
BunnyFl's firmware 13.42 check shows that the distortion, wide-near and
2D VR command queries give NOT_INITIALIZED for handle 0 before
sceHmdInitialize only when the title's compiled SDK version is below
4.50. From 4.50 the library first checks state it sets up when it
starts, which isn't measured, so handle 0 before initialization now
throws there, as it does after initialization and as all six queries
do on main. The SDK version is read from the process parameters, as
libSceHmd2 does.
guest_hmd checks handle 0 before initialization at SDK 0x044fffff and
0x04500000.
Source: https://github.com/boykopovar/AnyPS5/pull/9577#issuecomment-6343495321 (BunnyFl)
sceAgcCbSetShRegistersDirect and sceAgcCbSetUcRegistersDirect now accept
a list at any byte address because PPSA26344 passes lists at 1 mod 4.
The debt entry names that title evidence and states that how the console
reads such a list is not measured, so the host behaviour is not taken for
a measured console contract.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
FreeBSD 11's kern_connectat fails a nonblocking connect that is still
connecting with EINPROGRESS (36), a second one while it is still
connecting with EALREADY, and soconnectat fails one on a connected
stream socket with EISCONN. On Windows connect has reported
WSAEWOULDBLOCK as EINPROGRESS since 79e162110, but the poll test still
accepted EAGAIN (35) and the socket error numbers debt entry still
listed the gap. Require 36 in the poll test and cut the stale sentence.
Measured on a PS4 (firmware 10.71) for the review of the zone-shift
overflow: sceRtcFormatRFC2822 returns 0x80B50003 for UINT64_MAX with +1
and +1439 minutes and for 2^64 - 1 minute with +1, where the shifted tick
lands in year 1. The console returns 0x80B50008 when the shifted tick is
past 9999 as a signed value (2^63 with -1 minute) and otherwise 0x80B50003
when the source or the shifted tick is negative as a signed value
(2^63 with 0, 0 with -1); the tick after 9999 with -1 minute still formats.
sceRtcFormatRFC3339 does wrap on the console: UINT64_MAX with +1 minute
returns 0 and writes 0001-01-01T00:00:59.99+00:01, as the implementation
already does. Tests now cover both.
Written with AI assistance.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
d31e28308 made getdents and getdirentries throw on a name above the
one-byte d_namlen limit, and #8591 made getdirentries report the entry
index of its block through basep. The libkernel entry still described
ENAMETOOLONG and a basep of 0.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Measured on a PS4 (firmware 10.71) with an OpenOrbis homebrew that loads a
probe PRX calling libSceRtc; every output starts as a sentinel, so stores
and their width are visible.
- sceRtcSetTime_t takes the low 32 bits of its argument as unsigned seconds
and returns 0 for every value: -1 is 2106-02-07 06:28:15, 2^32 + 1 is
1970-01-01 00:00:01 and -2^31 is 2038-01-19 03:14:08. sceRtcGetTime_t
stays 64-bit (2106-02-07 06:28:16 is 2^32).
- sceRtcGetTime_t before 1970 and sceRtcGetWin32FileTime before 1601 store
0 and return 0x80B50008. Field validation still comes first and leaves
the output unchanged (1969-13-01 returns 0x80B50009).
- sceRtcSetTick and sceRtcSetWin32FileTime have no upper limit: the tick
after 9999-12-31 23:59:59.999999 is 10000-01-01 and the file time
2^64 - 1 is 60056-05-28 05:36:10.955161. sceRtcGetDayOfWeek accepts
years past 9999 (10000-01-01 is 6).
- sceRtcFormatRFC3339 returns 0x80B50008 when the zone-shifted tick is
outside years 1 to 9999. sceRtcFormatRFC2822 returns 0x80B50003 when it
is negative as a signed value and 0x80B50008 when it is past 9999.
The PS5-measured cases of GuestRtc.cpp that the probe repeated (RFC 3339
and RFC 2822 parsing, named and military zones, DOS time, TickAdd) give the
same results on the PS4. The three expectations changed here (SetTick past
9999, SetTime_t(-1), RFC 2822 at the limit with +1 minute) came from
commits that cite no measurement. Not yet confirmed on PS5.
Written with AI assistance.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sceNpCheckNpAvailabilityA, sceNpGetAccountLanguageA,
sceNpGetParentalControlInfoA and sceNpGetAccountDateOfBirthA return
SIGNED_OUT without writing, like sceNpGetAccountCountryA. The game
presence queries report OFFLINE. A null pointer or user id -1 is
INVALID_ARGUMENT, and the request forms refuse request id 0 and an
unknown request as sceNpSetGamePresenceOnline does. Signatures and
checks follow shadPS4 and Orbital; sceNpIsPlusMember keeps throwing.
Validate the configuration and resource information as
sceVideodecCreateDecoder does, map the configuration the same way and
report the memory and frame buffer requirements that
sceVideodec2QueryDecoderMemoryInfo gives for it.
