Each instance paced itself against the wall clock, so on a busy machine
the one that got less CPU fell behind, and anything a test times across
instances (a peer leaving, a matching timeout) came out differently with
the load: pair/vanish and pair/matching3 failed on macOS CI.
With --sync-file and --sync-count, instances on one machine share a file
of one slot per instance. CoreTiming ends a slice at every emulated
millisecond, publishes it there, and waits until every other instance
has reached it too, so they're never more than a millisecond apart. It
works for any number of instances. One that stops moving (a debugger
pause, a crash) is waited for 2 s and then left out until it moves
again, and one that blocks the CPU thread in real time on something the
others must do (the relay PTP connect) stands aside meanwhile.
Tools/adhoc_multi.py and adhoc_session.py use it (--no-sync turns it
off). The host's 1 s head start is gone: they wait for the server's
port instead. adhoc_multi writes each instance's output to a file, as a
full pipe would now stop the other instances too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The pair tests that pass go in tests_good, and each runs twice: through
the built-in server's relay and P2P. pair_known_failures lists the ones
that fail in only one mode. A P2P test that needs a loopback address
the machine lacks (macOS has only 127.0.0.1) is skipped with a warning,
and the run fails at the end with the command that adds it. The macOS
CI runners add 127.0.0.2. Each run gets a random port offset.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A real PSP encrypts its ad hoc link, so a sniffed capture of two consoles is
ciphertext. These decrypt and decode it (PDP/PTP/matching framing), to compare
what a game does on hardware against what PPSSPP sends at the sceNetAdhoc HLE
level - a debugging aid for hardware-vs-emulator divergence. Not needed to run
multiplayer (PPSSPP works at HLE and never encrypts) and unrelated to cross-play
(the link is console-authenticated).
The session key is ephemeral and not in the capture, but the data cipher has a
known-plaintext weakness, so a capture with some guessable plaintext is enough
to recover it - no console memory needed, and not a break of the key exchange.
The two fixed KIRK keys used already ship in ext/libkirk. docs/adhoc-packet-
decryption.md covers capturing, decrypting and recovery.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Games that show the system nickname in their lobbies can then tell the
instances apart. Also: press --times N --every S for menus that need
several presses, and start clears the last session's screenshots, which
could otherwise pass for this one's.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
until waits for a region of the screen to match a reference crop (or
a pixel its color), optionally pressing a button meanwhile, since
loading screens take a different time on every run and a press before
a menu is ready is lost. crop makes the references. log sets a log
channel's level and watch collects the log messages for a while.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Headless turns logging as a whole off without --log, so a channel
enabled with log.channel.set still logged nothing, and a client
watching the log events saw none.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Starts N headless instances of a game against each other, with or
without the relay, and drives them over the debugger: button presses,
side-by-side screenshots, frame rates, and scripts of those. Each
instance gets its own memory stick with a copy of the savedata, so
autosaves can't collide. docs/adhoc-multiplayer-testing.md describes the
setup, by hand too, and the network simulation options.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The u64 total length made both context structs 8-aligned, while the
game's context only has to be 4-aligned (hash/hash puts it in a char
array). On ARM32, GCC then copies buf with vld1.64 [:64], which faults
on such an address - on hardware too, not just under qemu as the
known-failures entry assumed. Store the length as two words; the layout
is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nothing ran the 32-bit ARM JIT in CI: android-arm32 only builds. The
arm64 runners can't run it natively on every host (some are Neoverse
V3, without AArch32), so this cross-builds armhf on x86-64 and runs it
under qemu-arm, the same way as riscv64 and loongarch64. The cross
entries now carry their triple and qemu binary in the matrix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A target time that had already passed, relativeUs 0 for one, failed
without a stop event, so a client waiting for cpu.stepping waited until
its own timeout. Now it doesn't run, and reports the stop right away.
A savestate loaded during the run kept the deadline's absolute time,
which belongs to the timeline the load replaced: a state from later in
the game stopped the run at once, one from earlier ran it for as long
as the gap. Now the deadline keeps the time it had left.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
game.state.save and game.state.load with a slot were limited to the
app's slot count (5 by default), which only means how many the UI shows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
--instance=N sets PPSSPP_ID, so each instance gets its own 127.0.0.N address
and MAC; headless left it at 0, which made its address 127.0.0.0.
--realtime keeps emulated time from running ahead of the wall clock. Only
display flips paced headless, and a test that doesn't draw never flips, so
two instances waiting for each other over the network timed out at once.
--adhoc-server runs the built-in ad hoc server.
Tools/adhoc_multi.py runs a multi-console pspautotest (X_host, X_client,
X_client2, ...) on one instance per role, with the built-in server or, with
--relay, a local aemu_postoffice in relay mode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lists the T() calls, settings page texts, config labels and choice lists whose
key isn't in en_US.ini. With --dynamic, it lists the lookups whose key is only
known at runtime instead, which it can't check.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Adds an arm64 simulator platform to the iOS toolchain. The simulator build
renders with Metal only: it doesn't link MoltenVK, and loads Vulkan at
runtime like the libretro core, so Vulkan reports itself unavailable and
the config falls back to Metal.
b-ios-simulator.sh builds the App Store variant and runs it on the booted
simulator, or on one of a given device type, for App Store screenshots of
devices we don't have.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The ffmpeg submodule held prebuilt libraries for every platform next to the
sources (about 1.5 GB checked out), and nothing uses those any more. Only the
needed sources are copied in, from the submodule's commit, along with every
header and template they include. The path is new, so checking out commits on
either side of this works.
It's also cut down to what PPSSPP uses: the H.264, AAC, MP3, ATRAC3 and ATRAC3+
decoders, the H.264 parser, the demuxers as before, and the FFV1 encoder, AVI
muxer and file protocol for AVI dumping. config.h takes its CONFIG_ values from
FFmpeg 3.0.2's own configure run with just those enabled. The raw H.264 demuxer
stays even though nothing opens raw H.264: probing PSMF data uses it to identify
the video stream's codec, which otherwise comes out as none.
On top of that:
- The files nothing links (configure would build them, but they're never pulled
out of the static library, going by the GCC link map) are left out: mostly
libavutil's crypto and hashing and the alternative DCT/IDCT/FFT versions.
- libswresample isn't built, now that nothing uses it.
- The H.264 decoder only does 8-bit video, which is all the PSP has. The SPS
parser rejects any other bit depth, and the DSP files no longer build their 9-
to 16-bit variants, which the linker kept because the init functions pick one
by bit depth at run time.
That's 160 files to compile instead of 277, and 391 of FFmpeg's vendored, about
5.8 MB. The README in ext/ffmpeg has what was done, the PPSSPP patches in the
copy, how to add a component, and the tradeoffs.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The 1.00 updater's DATA.PSAR is a version 1 archive, the only one there is,
and its blocks carry tag 0x06000000, which the PRX decrypter had no key for.
1.50's mesg_led.prx keeps its keys raw in a table after their tags, the
0x0E000000 one we already use among them, and the 0x06000000 one there
decrypts all of 1.00 (with KIRK code 0x49, following the other tags). Its
first record has no version text either, so the version now comes from the
firmware's own version.txt when that's missing.
Fix a zlib usage bug.
Headless --unpack-updater also exited 0 when every file failed, so now it
fails if any did.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Family Guy and The Dark Knight are coded as separate field pictures,
bottom field first. ffmpeg hands each pair back as one frame, and we
described it with the first field's SEI record only.
The UMD Video player (video_main_plugin.prx) shows such a picture a field
per vsync, and takes the field order from comparing the output times of
the picture's two records. With the second left empty it always chose
top first, so every pair ran backwards in time and the picture jittered
between frames. Now the pair carries both fields' timing, the top field's
record first.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The XMB's video player renders Motion JPEG AVIs with sceMpegBaseYCrCbCopyVme
and sceMpegBaseCscVme (JPCSP's names), which take planar YCbCr in two halves
rather than the ME's tiled layout. The copy sizes follow mpegbase_260.prx.
sceMpegBaseYCrCbCopy took the frame size from the destination descriptor,
which the same player leaves unset. mpegbase_260.prx reads it from the
source, in pixels.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
It wrote Key = Key, which for a key like PacketRelayHint that isn't its own
English text puts the key name itself in the UI of every untranslated
language. Now it takes the value from en_US.ini, falling back to the key.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A CD CHD is read through its track metadata: each track's frames are
padded to a multiple of four, a pregap is stored only when its type
starts with V, and audio is big-endian. A .chd that's a CD is a PS1 disc,
PSP discs being DVDs. The FileLoader-backed core_file the PSP CHD block
device used is now OpenCHDThroughLoader(), for both.
An .mds gives each track's start, pregap and length, and where in the
.mdf it's stored, with 96 bytes of subchannel data after each sector.
Anything between tracks isn't stored. A bare .mdf goes through its .mds,
which the game browser lists, and archives install with both.
MODE2/2336 cue sheets, which only had their sector headers made before,
are now tested too.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A cue sheet, a CloneCD .ccd, or a bare .bin, .img or .iso of a PS1 disc
now boots on the firmware's POPS, as the converted EBOOT.PBP popstation
would make from it. PS1EbootFileLoader builds the PBP header, PARAM.SFO
(the disc ID from SYSTEM.CNF) and the PSISOIMG header, TOC and block
index, and serves the disc image from the image's own files. Each block
of 16 sectors is a stored DEFLATE block: the manager won't read an image
of uncompressed blocks in its own format, but the converted PBP's inflate
takes stored ones, and with every block the same size, reads map straight
onto the files. Nothing is compressed or written anywhere.
Cue sheets can have any number of files, AUDIO tracks, INDEX 00 and
PREGAP, and a missing FILE falls back to the cue sheet's own name. A
2048-byte .iso gets Mode 2 Form 1 sectors made around its user data.
The EBOOT is mounted as the only file in host0:, through BlobFileSystem,
which now saves its open handles in savestates. The game browser lists
.cue, .ccd, .bin and .img, leaving out the .bin and .img files a cue
sheet or .ccd next to them describes, and a bare image opened directly
goes through the cue sheet that names it. GameInfoCache reads the title
and ID of these, and of real PS1 EBOOTs, from their PARAM.SFO.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A disc game or memory stick EBOOT started from the XMB used to be swapped in under the XMB's session,
so it ran without its PARAM.SFO, compat flags, game config, window title, savestates and so on.
Much of that is decided once, at boot (the memory map, the GPU backend, the shader cache), so
sceKernelLoadExec now asks the frontend for a reboot into the disc image or EBOOT (PSP_RequestBoot)
and stops. The app and headless boot it like a game from the game list.
The game keeps the XMB's launch args (CoreParameter::launchedFromVSH), so exiting boots the XMB
again, with the disc that's in the drive and those args, and the XMB comes back where it was without
auto-starting the disc. Homebrew keeps the disc that was in the drive. A disc whose executables are
all zeroes (the EUR Video Demo Disc's "game") still loads in place and fails back to the XMB.
In headless, the run carries on through the reboot, starting stopped under the debugger like the
first boot, since the reboot loses a pending cpu.runUntilTime.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The game list already did, but booting a video disc any other way (the command line, a file
association, dropping it on a running game, headless) tried it as a game and failed. The loader now
swaps in vshmain.prx with the disc mounted and auto-start on whenever it identifies a UMD Video
disc and the installed firmware can play it. GetUmdVideoVSHPath() moves from the UI to
Core/Util/PSARUnpack.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
With firmware 2.71 or later installed, picking a UMD Video disc boots the XMB with the disc in the
drive and UMD auto-start forced on for that boot, so the disc's menu comes up. Without one, the
info screen saying UMD Video isn't supported shows as before. The info screen's Play does the same.
A disc with both a game and video asks which to start. For the video, the XMB comes up with
auto-start forced off, since auto-start runs the game.
Older firmware can't play these: 1.50, 2.00 and 2.60 crash in video_plugin or paf with a video disc
in the drive, and 2.50 refuses it. 2.71 through 6.61 reach Bourne Supremacy's menu.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Post-shader framebuffers (for shaders at output resolution, and those
that keep the previous frame) are sized for the output they were built
for. Moving the game to a TV or back, or the TV changing size, left them
sized for the other one. SetOutputOverride now asks for a rebuild when
the output changes.
Also corrects the design doc: "auto" render resolution follows the
device's screen, not the TV.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Graphics > "External display" on iOS and Android. "Show the game" (the
default) is the TV mode, "Mirror the screen" leaves the TV to the system.
A change applies right away. It's disabled on backends that can't draw to
a second swapchain.
On Android 12 and up, a second screen that's part of the device (like a
dual-screen handheld's) isn't used, so the game doesn't move there by
itself.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A presentation display (HDMI or USB-C video out, some wireless displays,
or the simulated ones in Developer options) gets a Presentation with a
SurfaceView, whose surface becomes a Game output, when the backend is
Vulkan. Otherwise Android keeps mirroring.
The surface mustn't be touched after surfaceDestroyed returns, so the
output is removed with the new Native_RemoveOutputAndWait(), which
returns once the render thread has let go of it (or right away if the
graphics are down). The UI drawn on an output now leaves out the main
swapchain's pre-rotation, as the game already did.
"In menus on %1" gets the device name from Settings.Global.DEVICE_NAME.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Updates the submodule for video/mp4/avchold's ref3b1hrdsplice, recorded
on a PSP, which fails without the AvcDecoder fix on this branch: the pictures after
the splice never come out.
avchold and decodeend only test anything with Sony's mpeg and mp4
modules running, which headless never does for homebrew unless asked,
and which CI doesn't have. docs/pspautotests.md now says how to run
them (--disable-hle=0x110 and a memory stick with a firmware dump).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Swapchains other than the main one are VKRSwapchains, owned by the
render manager, each with its own surface, image views, framebuffers
and depth buffer, in the main swapchain's format so the backbuffer render
pass and pipelines work for all of them. A render step with no
framebuffer can name one.
They don't pace the frame: acquiring doesn't wait, and the pass is
skipped if there's no image. The frame's last submit waits on and
signals the semaphores of every swapchain acquired that frame, and each
is presented after the main one. When one goes out of date, it's
recreated at the start of the next frame.
The main swapchain's path is unchanged; frametests on Vulkan match.
VulkanContext::CreateSurface is the old per-platform surface code, now
also used for the extra ones.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The 1080p cap was meant to save Wi-Fi bandwidth, but an Apple TV 4K
only offers 3840x2160 over AirPlay, so the stream was 4K anyway, just
upscaled.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
An external display (like an Apple TV over AirPlay) becomes a Game
output, when the backend is Metal. The game is shown there full screen,
and the phone shows the touch controls with "Playing on TV" in place of
the game. The menus stay on the phone. With other backends, iOS keeps
mirroring.
Up to iOS 26 the external display scene comes from the Info.plist role.
iOS 27 ignores that, and only offers the scene to apps that register a
scene accessory for it, so we do that too. The registration is withdrawn
when the view controller shuts down, so it follows a backend change, and
a restart now goes to the app's scene rather than the first one.
An Apple TV 4K only offers 3840x2160 over AirPlay, so the stream stays
4K, but we render at 1920x1080 and let the layer scale it up.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The backbuffer is now a Swapchain, from GetPrimarySwapchain(), and
callers bind, measure and read it back through that instead of nullptr.
nullptr still means it too. Backends with one swapchain map the primary
to their old backbuffer path.
On Metal, CreateSwapchain() adds more, each with its own layer, depth
buffer and drawable. Passes merge per swapchain, and Present() presents
every drawable acquired that frame. The backbuffer is just the first
of them. Nothing creates a second one yet.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Not on tvOS, where the Metal backend hasn't been run yet. Existing
installs keep the backend saved in their ini.
The multi-swapchain plan now does Metal before Vulkan to match.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
It was left unresolved with the kernel mutex calls, as they all let the
flash drivers reach the NAND controller during a VSH boot. It's the
mutexes that do that: with only the FPL resolved, sceIdStorage_Service
allocates from its pool and every firmware's VSH in 1.50-6.61 boots to
the same XMB picture as without it. Resolving the mutexes still hangs on
0xbd101300.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Inserting and ejecting a disc there adds and removes the umd0:/umd1:/
umd:/disc0: mounts, and MetaFileSystem's state pairs sections with mounts
by position without recording which disc was in. A state could then fail
to load, or load one disc's open files onto another. The pause screen
already hid its save buttons for vshmain.prx; the save hotkeys, rewind,
the ImDebugger and the WebSocket debugger all went around that. The check
is now in SaveState itself, and rewind snapshots aren't taken. Verify
still runs, since it reads back with the mounts as they are.
A proper fix records the mounted image in sceUmd's state and remounts it
on load.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
