agc_shader_replay and agc_pm4_replay are built outside the tests output
directory, so the libc.prx they import loaded libstdc++-6.dll from PATH
and failed with 0xc0000139 when another MinGW, such as Git for Windows'
mingw64/bin, came before the toolchain. Copy the toolchain's runtimes
next to both tools after linking, as for agc_driver_visual_test.
sceAgcSetStaticBuffer, sceAgcGetStaticBuffer, sceAgcSetStaticQword and
sceAgcGetStaticQword read and write the six 16-byte slots at 0xfe0040000
that the driver already maps for the fetch shader table, through a new
AgcDriverStaticTable_nid_postfix accessor.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
Skipped nested/V# loads stay cacheable with empty poison; the chain that
still evaluates an unconditional pointer-pair load keeps two poison
records. Add the direct-root device case from Amoenus's #8272 validation.
WinLibs GCC 15.2.0 rejects the alignas(65536) static array that agc_driver_image_array_window_past_layer_limit added in #6772 (requested alignment '65536' exceeds object file maximum 8192), which stops a Windows build with every AGC execution test. Take a 64 KiB aligned 4 MiB span of a larger static array instead, as #6452 did for the indirect arguments block, and check its alignment at start.
The Windows -include of HostStandardStreams.hpp pulled in stdio.h before
libstdc++ could set __USE_MINGW_ANSI_STDIO, so long double formatting
bound to UCRT. Define the MinGW flag and include cstdio for C++.
The NetCtl ApRp calls, the SaveData debug calls, HttpDbgEnableProfile,
HttpDbgSetPrintf, PadEnableExtensionPort and VideoRecording Init, Term
and SendCommand return 0; ImeForTestFunction returns 0x80BC00FF; the Hmd
distortion work memory align and size are 0x400 and 0x20000 and
DistortionGetCompoundEyeCorrectionCommand returns 0x81110009;
AjmInstanceCodecType returns the instance id shifted right by 14.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
Host logging opens its own FILE on the saved descriptor. The guest FILE stays on the replaced descriptor, so guest printf, puts and fprintf still reach the pipe.
ContentBinder, CompanionUtil, ErrorDialog OpenDetail/OpenWithReport,
ImeDialog ForTestFunction, HttpCache SystemTerm/SendStatistics,
MoveTracker GetVersion/RecordStop, BluetoothHid and SystemGesture debug
versions and Mbus Term/AcquireControlList are a single constant return
in the firmware modules; these exports now return the same value.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
The clock, GPU CU count, activation and VR play zone calls return 0;
AcquireFb0, ReleaseFb0, the party voice and voice recognition calls and
LoadExecVideoServiceWebApp return 0x80A10002; LaunchUdsApp and
RequestToChangeRenderingMode return 0x80A10005. The firmware bodies of
these exports are a single return of that value.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
Follow FreeBSD 11's do_sem_wait and do_sem_wake (COMPAT_FREEBSD9/10) on
a struct _usem: a wait sets _has_waiters, returns at once while _count
is non-zero and otherwise sleeps on the semaphore's queue; a wake clears
_has_waiters when one waiter is left and releases one waiter.
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
This reverts the merge of #6932 (fce1fefe3b). Its "Install Xvfb" step fails on the self-hosted Linux runner, which has no xvfb-run and no passwordless sudo, so every Linux build stops before configuring.
On RDNA 2, an S# with TRUNC_COORD 0 rounds the scaled coordinate to 8
fractional bits, ties to even, and one with TRUNC_COORD 1 truncates it,
for point and bilinear filters. The S# decoder ignores the bit, and
AMD's Windows driver truncates for nearest samplers and rounds for
linear ones, so a TRUNC_COORD 0 point sample just below a texel edge
reads the texel below it. The code comment said the bit moves texel
selection by half a texel at most; the debt entry records the measured
rule instead.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Adrià