The test was registered only under if(UNIX) because it built its symbolic
link loop with ::symlink. On Windows it now makes the loop with
CreateSymbolicLinkW (unprivileged create) and exits 77, which CTest reports
as skipped, when the account cannot create symbolic links.
The public header audit found the three PS4 batch exports declared in libSceAjm's Api.hpp and in libSceAjm.native's, and the libSceAjmi run stub under its plain name. Keep them in libSceAjm.native and declare the libSceAjmi stub as sceAjmBatchJobRunBufferRa_nid_disambig1, as its definition is.
Co-Authored-By: Claude Haiku 5.5 <[email protected]>
A null batch address is refused whatever the size, so the accepted batch
of zero bytes is one at a real address, which is what TestConsoleEmptyBatch
starts.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A probe on a PS4 decoded exactly one frame and reported 0 for an ATRAC9
run job without RUN_MULTIPLE_FRAMES. Jobs read from a PS4 batch buffer
now do the same. Jobs from the PS5 builders keep decoding their whole
input, which FMOD in PPSA29141 relies on.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sceAjmBatchStartBuffer let a null batch of zero bytes through its
parameter check and formed the end pointer from it. A null batch now
returns INVALID_PARAMETER whatever the size, as the PS5 builders'
sceAjmBatchInitialize does; a batch of zero bytes still starts, runs
nothing and completes at once. A test covers both; neither was measured
on a console, which the debt entry says.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The gapless entry no longer says the one-frame behaviour is unverified,
since the probe measured it, and the libSceAjmi entry names the two
exports its file declares.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Review of the PS4 batch buffer change. The libSceAjmi stub of
sceAjmBatchJobRunBufferRa stays, declared with the _nid_disambig1 suffix
so that both libraries export the NID without defining one name.
sceAjmBatchStartBuffer reads the whole batch before running any job, as
probe batches on a PS4 showed: a buffer chunk of size zero is dropped
whatever its address, and a null address with a nonzero size refuses
the batch with INVALID_ADDRESS in the error record together with the job
chunk, the chunk's offset in the job and the job's return address, with
nothing of the batch run. Inputs and outputs share that rule. The
console-measured result flags move from test comments to the debt
entries.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
PS4 titles build AJM batches with the SDK's inline helpers, which call
sceAjmBatchJobControlBufferRa, sceAjmBatchJobRunBufferRa and
sceAjmBatchJobRunSplitBufferRa and start the buffer with
sceAjmBatchStartBuffer. The buffer belongs to the title, so it has to use
the console's chunk layout rather than this library's private records.
The builders write console jobs (job, return address, buffer and flags
chunks) and return the position after them. StartBuffer checks the
context, walks the chunks, skips inline data and runs each job as the
library's existing control, run or statistics job. Chunk headers are read
as the console's builders write them: bits 0-3 the kind, bit 4 a buffer's
output direction, bits 6-25 the payload; the other bits keep a reused
buffer's old bytes on a PS4. Malformed batches return MALFORMED_BATCH,
unknown chunk kinds and unimplemented control flags throw.
An ATRAC9 run with RUN_MULTIPLE_FRAMES reports which side ran out as a
PS4 does (0x8 input, 0x10 room, 0x18 both), and the ATRAC9 format reports
the frame's bitrate; both were measured on a PS4 with Knowledge is Power:
Decades (Wwise) batch dumps and probe runs. The vibration test's exact-fit
run expects 0x18 accordingly.
The throwing stubs in libSceAjm and the libSceAjmi copy of
sceAjmBatchJobRunBufferRa are dropped.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Fable 5.1 <[email protected]>
A DB_DEPTH_VIEW.MIPID above zero on a surface with DB_Z_INFO.MAXMIP zero selected no level and threw. Flashback sets it while it renders the six slices of a single-level 128x128 Z_X depth array, so the field cannot select another level; decode the first one as the console does.
Both tools merge with git merge-tree --write-tree, which needs git 2.38,
and pr_overlap's pairwise merges also pass --merge-base, which needs
git 2.40. On an older git, such as 2.34 on Ubuntu 22.04, they stopped
with git's merge-tree usage text and nothing said what was wrong. A
failing merge now reports the version it needs and the one found when
git is older; on a recent enough git the error is unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Implement pthread_attr_setcreatesuspend_np, pthread_resume_np,
pthread_resume_all_np and pthread_multi_np with the sce forms
scePthreadAttrSetcreatesuspend, scePthreadResume, scePthreadResumeAll
and scePthreadMulti, following FreeBSD 11.4 libthr and the firmware
13.42 libkernel_sys module. A thread created suspended waits before its
start routine, outside any cancellation point, until it is resumed.
sceKernelRaiseException throws on Windows for a thread still suspended
at creation.
A handler that read the mask and put it back with SIG_SETMASK stored its
own signal, which sigprocmask reports from the hold, in the process mask
and cleared the hold, so the signal stayed blocked after the handler
returned. pselect, setcontext and a siglongjmp inside the handler replace
the mask the same way. When the handler's own thread releases a hold, the
signal now gets back its bit from before the handler when the handler
returns, as FreeBSD's sigreturn puts back the mask from before it. A mask
another thread sets meanwhile is kept.
David