When sceAudioOutputBlocking has to wait and the wait fails at once
(interrupts or dispatch disabled, inside an interrupt), the firmware
returns the error but leaves the channel's waiting flag set, and the
channel can never be used or released again. Keep the error, drop the
rest: whether the channel was busy at that moment is timing, and a
small difference in ours could lose a channel for the rest of a game
where hardware wouldn't. No game can depend on losing one.
Savestates made while this was emulated have the flag cleared on load.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
From pspautotests intr/waits:
- sceAudioOutputBlocking sets the channel's waiting flag before its
event flag wait, and when that wait fails at once (interrupts or
dispatch disabled, or inside an interrupt) it returns the error
without clearing the flag. The channel stays busy from then on, and
can't be released.
- The SRC blocking output fails the same way even when a completion is
already there, leaving the buffer armed.
- After a block that had samples in it, the mixer DMA is still playing
it out, so a buffer arriving then isn't read early or restarts it.
- sceKernelVolatileMemLock only writes the fake address and size
through pointers that are there, instead of faulting on NULL.
intr/waits now runs to the end; one scheduling marker still differs, from
async IO timing.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
How to keep lldb and gdb from stopping on the faults the handler is meant
to catch, and the crash_*.prx tests as a quick check that it works.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Since 551e4cd0ab, headless without --graphics silently used the OpenGL
backend, since bSoftwareRendering defaulted to false. Under Mesa llvmpipe
on Linux/WSL, that hangs games early in boot (and test.py, which passes no
--graphics). The README already documents software as the default.
Also documents the traps hit while chasing this in docs/debugging.md.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
--disable-hle had no counterpart, which made "is this our fault or the game's?"
awkward to answer: the only ways to put our HLE back were a per-game config or
hiding the firmware, and neither works from a script - the setting is per-game
and the firmware gets found anyway. --force-hle takes the same bitmask and runs
our HLE for those libraries even where the real module is now the default, so
the same repro can be run both ways and the logs diffed.
Used it on the warnings left over in Tekken 6 under the real mpeg.prx. Three of
them appear identically with --force-hle=16, so they are the game's own and
match what the hardware answers: sceKernelChangeThreadPriority(-1) eight times
in a row (pspautotests/threads/threads/change says hardware returns
UNKNOWN_THID for -1 too), sceAtracAddStreamData on a released id with a zero
byte count, and a sceKernelDeleteMutex on a garbage id.
The fourth only happens under the real module and is ours. mpeg.prx sizes its
allocation through a scratch context in its own bss and calls
sceAudiocodecReleaseEDRAM on that one, while decoding through a different
context that never gets released - so the next movie's sceAudiocodecInit finds
a live decoder and replaces it. That is once per video on every game running
the real module, and it was a WARN_LOG_REPORT, so it would have reported from
everyone's machine. It is bounded - removeDecoder deletes the old one and Init
makes exactly one more - so it is an INFO_LOG now.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Also why gpu/signals/continue, jumps, simple and handlercalls fail, and the
one thing not to try on a real PSP.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
What SIGNAL and FINISH interrupts do and in which order, what enqueueing
checks, the pause window, what the sync calls wait for, and how
ProcessDLQueue() keeps the order right while executing lists in one go.
Also a note on gentest.py keeping a pipe open through usbhostfs_pc.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The test didn't compile: VERTS is captured by reference into testFormat, and
MSVC won't use a captured constexpr as an array bound. Make it static.
On arm64 it then failed on the UV prescale steps, by one ULP. The arm64 JIT and
the NEON handwritten decoders fuse the multiply-add, and the steps only match
that when the compiler contracts a * b + c - which clang does and MSVC doesn't,
in Debug or Release. Spell out which one happens instead of relying on it.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
When the firmware module a --disable-hle bit asks for is neither installed nor
on the disc, that library silently runs our HLE instead. For the emulator that
is the right thing; for a test tool it means the run measures something other
than what was asked for and says so only as one INFO line, which is easy to
grep past and easy to never see. It cost a round of wrong results here, where
the memory stick headless defaults to (beside the exe, not the app's) had no
firmware, so a comparison against the real mpeg.prx was quietly a comparison
against the HLE it was supposed to be measured against.
g_unavailableDisableFlags already records exactly which flags fell back, so it
just needed an accessor. Headless now names each one, prints the flash0:/kd and
memory stick it looked in, and fails the run.
Only an explicit --disable-hle binds. sceMpeg and sceMp4 are LLE by default, and
falling back is the correct and expected behaviour wherever no firmware is
installed - making that fatal would fail every run on such a machine.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Running a commercial game through headless to measure something has four ways to
report a clean pass for a run that proved nothing, and they compound: --log is
needed before anything is printed (and it goes to stderr, so grepping forces the
streams together), the memory stick defaults to one beside the exe rather than
the app's, which means no installed firmware and a silent LLE-to-HLE fallback,
an unrecognised parameter then hides in the merged output, and a count of zero
errors reads identically whether the code ran or never got there.
All four of these cost a round of wrong results while investigating ME memory.
Also records that the line-ending check has to be done on the bytes, since Git
Bash's grep normalises them: the obvious `grep -c $'\r$'` for a stray LF reports
every file clean, which is how a handful of LF lines got into two CRLF files
here without the usual whole-file-rewrite tell in git diff --stat.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The legacy Android build has a unit test executable of its own, so a new file in
unittest/ goes in three build files rather than the two the docs named. Missing
the Android one builds fine everywhere it is convenient to try and fails only on
Android CI, which is what happened here - so both AGENTS.md and building.md now
say three, and which one is easy to forget.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The Ant build hasn't been runnable since the SDK dropped tools/ant/, so build.xml,
custom_rules.xml, ab-ant.sh, ant-build.bat, project.properties and
proguard-project.txt all go. Plus some orphans: buildassets.sh (nothing called it),
build.sh (still referred to "phoenix"), symx86.cmd (x86 isn't in APP_ABI) and
README.TXT (Eclipse import instructions).
ab.sh and ab.cmd no longer copy assets into android/assets - only the Ant/Eclipse
packaging ever read that directory. Gradle packages the repo-root assets/ directly
and ndk-build doesn't look at assets at all, so the stray copy only made it unclear
where the APK's assets come from. Noted how that actually works in docs/building.md.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The MSBuild examples passed /p:Platform=x64 and the run lines pointed at
Windows/x64/..., so following them on an ARM64 machine produced an x64 build -
which then runs anyway under emulation, so nothing looks wrong. It is slower
than the native build, it isn't the code ARM users get, and a benchmark taken
from it measures the emulator: the colour conversion benchmark this was noticed
on reads 200 MPix/s emulated against 300 native.
The examples now say <platform> rather than either value, so there is no default
to follow and the machine has to be looked up. Also note that
$PROCESSOR_ARCHITECTURE describes the shell, not the host, and says AMD64 from
an emulated shell.
test.py searched only Windows\x64 for the headless binary, so on Windows-on-ARM
it would silently test an emulated build; it now looks for the native one first.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
--timeout was wall-clock seconds, which is what CI wants but not what you want
when the question is whether the game has had long enough to get somewhere: a
heavy scene runs many times slower than real time and a near-idle one much
faster, so the same budget means very different amounts of game time. Booting a
firmware VSH is a good example - 10 emulated seconds is about 25 real ones on
6.61 and about 7 on 2.00, and judging those two by the same wall-clock number
makes a working shell look stuck.
Both limits can be set at once and whichever is reached first ends the run,
which also says which one it was. --timeout still works as the old name for
--timeout-wall. The IsDebuggerPresent() exemption stays on the wall-clock check
only; the emulated one doesn't need it, since sitting at a native breakpoint
burns no emulated time.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Another pass looking for gaps, all of it now recorded by audio/blocking/channels
and audio/blocking/oneshot.
sceAudioOneshotOutput was the last unimplemented entry in the module, returning
"library not linked" to anyone who called it. It plays one buffer on a channel it
never reserves, so the channel frees itself when the buffer runs out, and its
argument checks are their own set: any positive sample count, aligned or not, no
upper bound, a negative volume rejected rather than skipped, and no busy check at
all. No game is known to use it; it is implemented because tracing it turned out
to be cheap, not because anything needed it.
The channels test confirms four behaviors: sceAudioChReserve(-1) skips a released
channel that is still playing while a reserve of it succeeds, a second channel
joining a running mixer doesn't lose a block unlike a first, mono counts down in
the same 64-sample steps over the same time as stereo, and the panned blocking
output has no extra delay on the high channels.
Also: the SRC resampler now interpolates into the next buffer at a buffer join
instead of holding the last sample, since the codec reads the two descriptors as
one stream. That only shows up at non-native rates and no test can see it.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The nine channels were one array of one struct, but only index 8 ever used the
two DMA descriptors, the played/fraction position and the waiting-thread vector,
and only 0-7 ever used the buffer slot - so eight copies of a std::vector sat
there dead. They are two different pieces of hardware and now they are two
structs: AudioChannel for the eight the mixer walks, and a single AudioSRCChannel
for the one the codec reads directly.
Found while rereading it:
- __AudioUpdate can now run from inside an output call, and it reschedules when
it wakes a thread. A context switch there lands before the syscall writes its
return value, so the caller loses it and the woken thread gets it instead.
Deferred with hleReSchedule when a syscall is in flight. The same fix applies
to the release path, which had the problem before this branch existed.
- A finished SRC buffer whose waiting thread had given up dropped the completion
entirely, so the next caller waited a buffer too long. Now it walks past dead
waiters and banks the completion if nobody is left.
- An output with a null pointer changed the channel volume, where the hardware
returns before touching it.
- A busy channel came back as an error from the Output2 path and as debug from
the mixer path. It is an ordinary answer a game polls on, so both are debug
now; the old behavior filled the log with 18k error lines in a minute of F1
2009.
- AudioChannel::reset had no callers left: releasing a channel with a thread
parked on it is refused, so there is nobody to wake.
- The two routing modes are globals and were written once per channel. They get
their own small savestate block.
Savestate: the sceAudio section goes to 3, and the SRC channel gets a section of
its own. States from released builds carry it as a ninth channel record, which is
read and discarded.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
I deduced that this was the case, and attempted implementing this path long ago,
but I could never quite get it to work in all games. Set Claude on a quest to
research and implement it, and lo and behold, it works. A bit sobering.
Fixes#12888 and likely more. Additionally, audio latency is likely slightly
improved overall, and memory usage is down by 4.6MB.
Claude says:
The blocking output calls are not a queue that callers line up behind. Each
mixer channel holds exactly one buffer and at most one parked thread; a second
thread arriving while the first is waiting is told the channel is busy and is
expected to skip its turn. The Output2/SRC channel holds two DMA descriptors and
refuses a third caller outright, without waiting at all.
We blocked everyone instead, so a game running a movie thread and a sound-effect
thread over one output made the two alternate - a frame of movie audio, a frame
of effects silence - and the movie played at half rate. That is #12888, seen in
F1 2009 and Colin McRae: DiRT 2. With this, the movie thread keeps the channel
for the whole cutscene and the effects thread is refused, which is what the
hardware trace shows.
The driver also never copies a buffer on the way in: it stores the pointer and
its mixer walks it forward 64 samples at a time out of the game's own memory.
Modelling that fixes#20095 as a side effect, and drops the 4.6MB of per-channel
sample rings we were carrying. The mix event is re-phased to the moment a DMA
starts, since the mixer thread outranks its caller and gets a block in before the
output call returns.
Along the way: the two rest-length calls differ after a null-pointer output,
sceAudioChRelease reports not-reserved rather than not-init,
sceAudioChangeChannelConfig validates the format,
sceAudioChangeChannelVolume validates nothing, and sceAudioOutput2ChangeLength
takes a range of 17..4111. Details in docs/sceAudio.md.
Savestates: AudioChannel goes to version 4. Older ones stored mixed samples that
can't become a pointer and a position again, so they load with the pending audio
dropped and any parked threads released.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Record the fifth cause (the rejected impose parameter), correct the counts now
that 3.80/3.90 and 2.6x-2.8x are covered, and note that 1.50 through 2.50 never
touch the encrypted XMB index at all.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The old claim of "all of them" measured that the shell renders, which several
versions do while showing only the wallpaper or a full-screen error. Record what
each version actually does, the four causes that were in the way, and how to
find the next one of each kind - including that --timeout is wall-clock, so a
slow-drawing shell can look stuck when it isn't.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Backwards from 5.01, one release at a time, against all 39 versions that ship
on a disc plus the download-only 6.61. Nothing here is an offset - it's almost
entirely Sony renumbering the kernel *_driver NIDs, which sends an import we
mean to HLE into the real firmware module instead.
Four more NIDs each for sceRtc_driver/sceRtcSetAlarmTick and
sceHprm_driver/sceHprmReadLatch, covering 1.50 up. The rtc one is what parked
every thread on a SceSysconSync semaphore; the hprm one runs once a frame, so
unresolved it was most of the boot log. Also sceImposeGetParam/sceImposeChanges
(1.50 - 2.xx) and sceKernelLoadModuleVSH (1.x, which is how the shell loads its
own plugins - unresolved it got module id 0 and StartModule failed).
sceRtcIsAlarmed had to be implemented too; it returns 0, as in JPCSP. As a null
entry it returned LIBRARY_NOT_YET_LINKED, and the 3.0x-3.5x VSH read that as
"ask the hardware instead" and went back to blocking on syscon.
Two structural findings:
- Up to 4.05, scePaf's heap allocator is a separate heaparea1.prx that paf
imports as scePafHeaparea. Load it when it's there. Its pool pointer needs
the same pre-fill paf's does, at gp - 0x7FCC rather than gp - 0x7E88.
- 1.50's vshmain.prx declares no module attributes at all - PSP_MODULE_VSH_MODE
only appears from 1.52 - so the whole VSH bootstrap was being skipped. Accept
the module name too.
Co-Authored-By: Claude Opus 5 <[email protected]>
The blocker below 6.60 wasn't offsets, it was that Sony renumbered the kernel
*_driver NIDs between versions. A function we HLE under its 6.6x NID is a
stranger on an older build, so the import lands in the real firmware module
instead - and that's where it goes wrong:
- sceRtc_driver sceRtcSetAlarmTick. Without the HLE the VSH's alarm call ran
the real rtc.prx, which called on into syscon.prx and blocked forever on a
SceSysconSync semaphore. That was the whole "stalls with every thread parked"
symptom; the tell was a fourth SceSysconSync waiter a healthy boot lacks.
- sceHprm_driver sceHprmReadLatch, called once a frame - so before this an
older firmware's 12-second boot logged ~20000 lines of one unresolved import.
Three extra NIDs each, found by disassembling the module from both firmwares
and matching on the address of the user-mode export whose NID never changed
(sceRtc/0x7D1FBED3, sceHprm/0x40D2F9F0).
5.55 additionally needed two PRX decryption keys we didn't have (0x4C941AF0
and 0x4C941BF0) - without them none of flash0:/kd decrypted and the shell came
up with no drivers behind it at all.
Checked one release at a time against every version that ships on a disc, plus
6.61. 4.05 and below still die on a null write inside vsh_module.
Co-Authored-By: Claude Opus 5 <[email protected]>
6.60 ships byte-identical paf.prx and vshmain.prx to 6.61 - all 6338 + 669
functions disassemble the same - and boots to an interactive XMB, so let
FirmwareVersionSupportsVSH accept it. That matters because no UMD carries 6.61
(it was download-only), so 6.60 is the best a disc-installed firmware can be.
The two module patches were hardcoded offsets from the module base applied to
any module of the right name, which is quietly wrong on any other build:
- The scePaf heap arena slot moves with every build (0x18CCD8 on 6.00 through
0x18D728 on 6.60/6.61) but sits at gp - 0x7E88 in all of them, so find it
that way. On its own this turns an immediate SIGSEGV inside scePaf into a
clean stall on 6.00 through 6.39 - they still don't reach an XMB, they get
stuck in sceVshBridge_Driver instead.
- The vsh_module alarm-category offset has no such anchor, so check the word
there is the one the patch was derived from. On 6.20 and 6.00 it's ASCII
string data - the unconditional write was corrupting a string table.
Also resolve the per-model kernel drivers (memlmd, loadexec, wlanfirm) to the
model being emulated. They were asked for as _01g, which a firmware unpacked
for a single model doesn't have - and our own updater unpack defaults to 02g.
Co-Authored-By: Claude Opus 5 <[email protected]>
A tool nobody knows about is a tool nobody uses. AGENTS.md gets a short section
pointing at it, plus the two things most likely to be got wrong when reading
the output: that a function's arity can't be inferred from the registers it
reads, since MIPS code passes arguments through untouched, and that a finding
is worth much more when the comment says which module it came from.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
PPSSPP decided whether a caller was privileged with hleIsKernelMode(), which reports whether the
syscall being executed is itself a kernel-only export. That's a different question from the one
the hardware answers: on a PSP the privilege belongs to the calling module, and a kernel module
reaches sceKernelCreateTlspl through the ordinary ThreadManForUser NID like anything else. So a
kernel module asking for partition 1, 3 or 4 got ILLEGAL_PERM where a real PSP hands it over,
which the new threads/tls/kernel/partition test shows directly.
BlockAllocatorFromID now also accepts a caller whose thread belongs to a kernel module, via a new
__KernelCurThreadIsKernelMode(). It checks the thread's own attribute first and then the owning
module, because a kernel module's main thread isn't necessarily flagged kernel - the attribute
comes from PSP_MAIN_THREAD_ATTR, which needn't set it. That mirrors how sceKernelCreateThread
already works out allowKernel.
This only ever widens access, and only for threads belonging to kernel modules, so games are
unaffected - they run in user modules and see exactly what they saw before.
threads/tls/kernel/partition now records the sweep from a kernel module, which settles the range
question the user-mode recording couldn't: privilege changes the permission check, not the range.
Partitions 1, 3 and 4 are ILLEGAL_PERM from user mode and fine from kernel mode, while 7 and up
are ILLEGAL_ARGUMENT either way. So the check goes back to a plain 1-6 for both, and the
kernel-mode carve-out from the last commit - which would have let 8 and 9 through - is gone.
The hardware doc gains a section on kernel-mode tests: what COMMON_KERNEL does, why the stock
crt0 makes a kernel PRX unloadable, which libraries can't be imported, and how much room there
actually is in the kernel partition.
Every test directory builds under pspdev GCC 15 again, so gentest.py no longer
aborts on a neighbour's compile error before it reaches the PSP. No .prx was
regenerated, so the suite behaves exactly as before - 319/319 still pass.
Also updates the hardware doc: the "several tests don't build" workaround is
gone, and it now records what rebuilding a .prx actually costs (64-bit time_t
changes what rtc/convert tests), that host0: differs per host OS, and that
PRXs stay resident so you need a reset between runs.
We had a doc for running the existing tests against headless, but nothing on
the other half - bringing up PSPLink and usbhostfs_pc, what gentest.py does,
and how to get an .expected out of real hardware. Write that down, including
the parts that cost time to rediscover: usbhostfs_pc's working directory is
host0:/ so it has to start in the pspautotests root, gentest.py makes the
whole test directory and several old tests no longer build under pspdev's
GCC 15 (use -k), rebuilding a .prx with a newer toolchain balloons it, and
host0: is not FAT so anything testing FAT semantics needs ms0:.
Also adds the io/shortname test the doc uses as its worked example. It stays
in tests_next: hardware preserves the case of d_name where we uppercase it,
and appends ~1 to the short name of anything that isn't already valid
uppercase 8.3 where we only do that on a collision.
threads/tls/create moves to tests_next as well. It's collateral from the
submodule bump - upstream 1dcefeb regenerated its .expected on a PSP with
less free memory, so allocations at 1MB and above now expect failure, and
partitions 8 and 9 now expect 800200D2 where we return 800200D1.
A .sprx from one of these packages is an NPDRM "\0PSPEDAT" container: a
0x90-byte header, then an ordinary ~PSP PRX. The loader only ever saw the
EDAT magic and gave up with SCE_KERNEL_ERROR_UNSUPPORTED_PRX_TYPE.
Step over the header, then derive the key the PRX inside is really
encrypted against: sceNpDrmGetFixedKey() over the content ID, XOR in the
licensee key the game handed us through sceNpDrmSetLicenseeKey(), then AES
under a module key that had to be added. Both halves of that were already
lying around unused - sceNpDrmGetFixedKey() had no callers at all, and the
licensee key was being kept and never read.
The rest of it is a fixed XOR that the PRX header's decrypt_mode selects
rather than its tag, so it's applied on the mode the way JPCSP does it and
the tag table is left alone - tag 0x407810F0 carries no seed of its own
there either, so ours was never wrong about it. pspDecryptType5() already
had a slot for both XORs; no new decryption logic was needed.
Decryption is only half of it: these modules are KL4E-compressed rather
than gzipped, so they also need Core/Util/KL4E.cpp, which is already there
for the firmware modules that use the same compression. With both halves
Shiren 4 Plus loads its one big .sprx and runs. God Eater 2 needed one
further fix that isn't in this commit - the type-B relocation bug in
ElfReader::LoadRelocations2, issue #8075 - and then plays.
docs/pkg_notes.md has the container layout and the key derivation.
Co-Authored-By: Claude Opus 5 <[email protected]>
An installed update silently replaces what the game boots, so the info
pane now says when there is one - version, size and where it lives - and
the context menu offers to remove it again.
Removing takes the whole PSP/GAME/<DISC_ID> folder when the update is all
that's in it. When a digital game shares the folder, only PBOOT.PBP goes,
since deleting the folder would take the game with it and nothing records
what the install wrote. The confirmation names the exact path either way.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_018izZ1mGTWhz2RqeudqsDQR
Opening a .pkg now offers to install it, the way a .zip does - see the new
InstallPkgScreen, which shows what the update patches, what it'll take up
on disk (exact, since package contents aren't compressed) and where it's
going. The package's PS3-style USRDIR/CONTENT wrapping is stripped so the
files land where the PSP expects them, in PSP/GAME/<DISC_ID>.
Booting a disc then looks for PSP/GAME/<DISC_ID>/PBOOT.PBP and boots that
instead of the disc's own EBOOT, leaving the disc mounted - so the update
overrides the files it ships and the disc supplies the rest. The update's
DISC_ID has to match; a DISC_VERSION mismatch only warns, since updates do
get used with slightly different dumps in practice.
Verified against the whole corpus: every one installs, and the
digital NP* update/base-image pairs that could be assembled all boot the
patch rather than the disc's executable. That includes Super Robot Taisen
Operation Extend from a real NPUMDIMG EBOOT.PBP, which settles that
ISO.BIN.EDAT does not re-key the PBOOT - a digital title's patched EBOOT is
encrypted exactly like a UMD one. On the UMD side, the patched
LittleBigPlanet reads PATCH.ARC out of the install alongside the disc's own
archive.
The DISC_VERSION warning turns out to be load-bearing: many of the pairs
mismatch, because the dumps in circulation are later disc revisions than the
updates were built against. docs/pkg_notes.md has the numbers, and the one
thing that doesn't work - PGD-wrapped .sprx modules, which
sceKernelLoadModuleNpDrm can't load.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
A standalone Python reader for the .pkg containers Sony shipped PSP game
updates in: it prints a package's header, metadata, both PARAM.SFOs and its
item table, and extracts the payload. Nothing in PPSSPP calls it - it exists to
work the format out and to have a second implementation to check the C++ one
against, the way Tools/ already holds a few other one-off analysis scripts.
docs/pkg_notes.md is what it was written from: the header and item table
layout, the two AES-CTR keys a single package mixes, and the detail that trips
up a first attempt - which key applies is per item, not per package, so a
reader that picks one produces garbage filenames for most of a package while a
few entries decode perfectly.
Co-Authored-By: Claude Opus 5 <[email protected]>
The tracer records every basic block the CPU executes and can write the
instruction stream out to a file, which is the tool you want when something
corrupts state and the question is "what actually ran just before". It was
only reachable from Developer Tools in the UI, so a scripted session had no
way to turn it on, and reconstructing the same thing from log-only breakpoints
means guessing what to watch before you know what happened.
Five events: cpu.tracer.start/stop/flush/clear/status. start takes the two
buffer sizes and clears the JIT cache by default, because blocks compiled
before tracing was on don't carry the LogIRBlock instruction the tracer feeds
on - without that a hot loop compiled earlier simply never appears. The trace
ring is cyclic, so a finished recording holds the last maxTraceSize blocks:
start it, run into a crash, and the tail of the file is the instructions that
led there.
Only the IR cores drive the tracer, so start refuses on the others and says
which core is loaded rather than recording nothing; status reports the same
thing as `supported` so a client can tell that apart from "nothing executed".
Everything that mutates tracer state goes through Core_RunOnCPUThread, per
docs/DebuggerThreading.md.
Co-Authored-By: Claude Opus 5 <[email protected]>
The command splits $1/$2/$3 off the invocation positionally and then states them
as fact, so calling it with a sentence rather than `/add-string <Section> "<Key>"`
yields three arbitrary words - and nothing downstream notices. Ask for them to be
checked against en_US.ini first, and to re-derive the real section and key from
the request if they don't hold up.
Also spell out, in both the command and docs/translations.md, that the en_US line
in the scratch file overwrites en_US.ini like any other language, so for an
existing key it has to match the current English text exactly - otherwise it
quietly rewords the string every other language was translated from.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01GgACRqkQNpfJQ4fjwyoEup
Records the recurring bug classes, the things that look wrong but are verified
correct (so they don't get re-flagged), what was deliberately left alone, and
how far the review actually got.
