Prospero Themes (PPSA90210) imports nmount (3YnXspQNyw8) from libkernel
and fails to load without it:
"./app.elf: symbol lookup error: ./app.elf: undefined symbol: 3YnXspQNyw8".
Mounting file systems is not supported, so it throws if called, like
execvp does for executable replacement. The title opens its theme
library without calling it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Sovthwyu4H63RMbRVha2Sy
A change to build.yml ran the relinker, register-context and write-watch jobs whatever it touched. The patch's lines are now mapped to the job sections of the workflow at the head, and a test job runs only if its own section, the eligibility job or the shared part outside jobs changed; without a patch every job runs as before.
Co-Authored-By: Claude <[email protected]>
PSFlyCast (PPSA99247) imports pthread_condattr_setpshared (3BpP850hBT4)
from libkernel and fails to load without it:
"./app.elf: symbol lookup error: ./app.elf: undefined symbol: 3BpP850hBT4".
The attribute stores PTHREAD_PROCESS_PRIVATE (0) or
PTHREAD_PROCESS_SHARED (1); a null or uninitialized attribute and any
other value return EINVAL without changing it, following FreeBSD 11
libthr thr_condattr.c and pthread_mutexattr_setpshared here. Sharing
across processes is not modeled, as for mutexes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Sovthwyu4H63RMbRVha2Sy
A device that imports shared memory through host pointers, such as lavapipe, imports the unaligned range that way, and a host-pointer import has no lead.
A unit shadow published its units into the import's buffer at the guest offset from the import's
base, without the import's lead, so a sparse import that starts below its range (HostImport::lead)
got the units lead bytes before their place. The shadow keeps the lead with the import's base and
buffer, and the shadow verify reads the guest bytes at the same place.
agc_driver_sparse_dma_buf_import publishes a unit shadow of its unaligned import and checks the
unit's guest bytes and the bytes around it; without the lead it fails there.
A sparse import binds whole 64 KiB blocks, so a range whose memfd offset
is not a multiple of 64 KiB fell back to one UDMABUF_CREATE and, above
udmabuf's size limit, to copies. The sparse buffer now starts at the
block below the range: HostImport::lead holds that distance, the
import's device address points at the range itself, and every offset
into the import's buffer goes through HostImportOffset.
Queue builds are workflow_dispatch runs on main and never saved their ccache,
so each one restored the last push build's entry and recompiled everything the
queued PRs and the merges since then touched (32 min vs 6 min warm). They now
save under their own prefix, restore it first and fall back to the push
entries; push builds still restore only push entries, so main never starts
from a tree that held unmerged PRs.
Co-Authored-By: Claude <[email protected]>
The [dispmove] line printed refused 0 next to a rebuild count that came mostly from PrepareDispatch dropping the hit. The line now reports the refusals at stage A and at record separately, each with its reasons.
EmitBdaAnchors declared the u64 type before checking for anchors, so every
program got a 64-bit integer type, and programs without the Int64 capability
failed SPIR-V validation. It now returns before declaring anything when the
program has no anchors. A compute program without BDA reads is checked to
declare no 64-bit integer type and no Int64 capability.
AI-assisted: yes
select(c, a, select(c, b, d)) is select(c, a, d), and
select(c, select(c, a, b), d) is select(c, a, d). The recompiler leaves
these behind when a register is written more than once under the same
lane mask and when the f32 NaN selection tests the same source twice,
and nothing folded them. The constant folder now takes the inner arm
the outer condition already decides, then folds a select whose arms
became equal.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Goosebumps Terror in Little Creek (PPSA27305) aborts in gameplay with
"unsupported color format 16 number type 0 component swap 2": it renders
into a COLOR_5_6_5 UNORM target. The target now renders into
R5G6B5_UNORM_PACK16. AMD keeps X in bits 0-4 and Z in bits 11-15, and
Mesa's ac_translate_colorswap programs SWAP_STD_REV for a surface with
red in bits 11-15, so with SWAP_STD_REV the exports are written as they
are and with SWAP_STD they are written with X and Z exchanged, as for
1_5_5_5. The alternate swaps and other number types keep throwing.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
With #6014 merged, the first draw of METAL GEAR SOLID: Peace Walker's
(PPSA18216) opening movie stops on "the depth target's sample count
differs from the rasterizer's" (mrfelek on #6196). The state he reported
is a 4x PA_SC_AA_CONFIG, MSAA_ENABLE clear, a one-sample 1440x810 depth
target and no color target; on main alone it throws earlier, on "a depth
target". Like the one-sample color targets of bb877791a, such
a depth target, with or without one-sample color targets, is now drawn
as a one-sample draw at the pixel center, before the multisampled-target
check and instead of the sample-0 path, which needs MSAA_ENABLE.
Points, lines, alpha-to-coverage, a sample-mask export and DB_EQAA Z
interpolation or overrasterization throw on this path, and so do
PS_ITER_SAMPLE and ANCILLARY or SAMPLE_COVERAGE pixel inputs when the
pixel program runs. A skipped pixel program is replaced by the null
program, which uses neither.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Prevent reached null conversion destinations from crashing sscanf_s and vsscanf_s. Return EOF before writes and retain earlier assignments. Add regression coverage through both guest ABI entry points.
AI-assisted: yes (OpenAI Codex). Windows WinLibs GCC 15.2 Release: production baseline crashes with the new regression; fixed shared libc passes guest_scan_count, guest_formatting, guest_wide_formatting and guest_strings (4/4). Conventions pass. Full all-target build was deliberately stopped after 83s with 3691 pending targets, not completed. No title or console run; constraint-handler invocation remains unimplemented as documented. No performance or coverage-percentage claim.
Adds tests for nine implemented libSceRudp functions that previously
lacked coverage:
- sceRudpActivate (returns 0 always)
- sceRudpEnableInternalIOThread (NOT_INITIALIZED before init, 0 after)
- sceRudpSetEventHandler (returns 0)
- sceRudpBind (NOT_INITIALIZED, INVALID_CONTEXT_ID, 0 on success)
- sceRudpSetOption (NOT_INITIALIZED, INVALID_CONTEXT_ID, 0 on success)
- sceRudpInitiate (NOT_INITIALIZED, INVALID_CONTEXT_ID, CONN_REFUSED)
- sceRudpGetContextStatus (NOT_INITIALIZED, INVALID_CONTEXT_ID, fills
status with 0, null status allowed)
- sceRudpWrite (NOT_INITIALIZED, INVALID_CONTEXT_ID, CONN_RESET)
- sceRudpRead (NOT_INITIALIZED, INVALID_CONTEXT_ID, CONN_RESET)
Each assertion mirrors the exact implementation in Export.cpp. Error
codes and return values are the constants hardcoded in the source.
No console measurement needed — these are pure offline behaviors.
AI-assisted: yes
A v_mul_f32 in a FLOAT_MODE that flushes f32 denormals emits the
nearest-even tininess check, the NaN result selection and then a
result flush. The check already turns every product that rounds below
2^-126 into a zero of its sign, so the flush can't change what it
keeps. DenormalFlushEliminator now proves that: a select arm guarded
by the tiny-product check of the same denormal-free factors, an arm a
select takes only when its own bits are NaN, and an and with a mask
that has no significand bits are denormal-free.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Both jobs checked out without a ref, so a dispatched build compiled and
tested main instead of the selected pull requests, although the gate
starts them because a pull request touched their files. They now check
out the gate's main commit and merge the pull requests with the same
step as the relinker, Linux and Windows jobs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A guest path is resolved by appending its components to the host root or
to a mount alias directory through std::filesystem::path. On Windows a
component such as "C:" that came first in the appended part was parsed as
a root name, so operator/ replaced the host directory instead of
appending: "/savedata0/C:/Windows/win.ini" and "/app0/../C:/Windows/win.ini"
resolved to "C:\Windows\win.ini", and "/C:/x" lost the component and
resolved to "<root>\x". A colon is an ordinary character in a FreeBSD
file name, so those names belong under the mount.
Guest paths are now split into component names and joined to the host
directory as text. Explicit native drive paths are still passed through.
On NTFS a name containing ":" still opens an alternate data stream; the
libc debt list now records it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
MayEvaluate and the evaluator track the first-lane reads they are inside, so a read reached again through a loop-carried phi or a constant slot is not evaluable instead of recursing.
A bind of slice 1 or later of a depth array read the slice's HTILE but took
its clear only when all its words were equal, because the HTILE word layout
of those slices is not modelled. Neither step needs the layout: the clear
is taken when every word of the range has ZMask 0 (a check over the set of
words), and the store-back rewrites each word in place from its own bits.
The value the surface is cleared to is DB_DEPTH_CLEAR, which the DB's depth
test reads for every cleared tile whatever its Min/Max Z (HTILE_Z_CLEAR_REG
in Mesa's src/amd/common/ac_descriptors.h), the same rule slice 0 already
follows.
PPSA05684 resets the four slices of its 2048x2048 D16 sun cascade array
every frame by writing ZMask 0 words with Min/Max Z 0 or 1 into their HTILE.
Slices 1 to 3 kept the previous frame's depth, and the main menu lost its
volumetric sun shafts and sunlit foliage.
agc_driver_depth_fast_clear_submit: slice 1 with words alternating 0x0 and
0xfffffff0 takes the clear, stores them back expanded and leaves slice 0's
HTILE alone. The refusal case keeps a word left expanded and is renamed.
AI-assisted: yes
agc_driver_sample_multisample_target draws into a 4x target and reads
its samples back through a multisampled storage image. On a device
without shaderStorageImageMultisample, such as an Intel Arrow Lake iGPU
on ANV, the recompiler refuses that shader ('storage image multisampling
is unavailable on the target device') and the test failed. It now
skips there, as the recorder's storage multisample checks already do.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
