## Link to GitHub Issue or related Pull Request, if one exists
n/a
## Description of change
Fix hang when a DLL using the SDK calls `log_fatal` (deadlock)
Fix spicecfg crashing if API server fails to bind (make websocket
teardown conditional)
## Testing
## Link to GitHub Issue or related Pull Request, if one exists
developer request
## Description of change
Expose coin insert API over the SDK
## Testing
see sample sdk code
## Link to GitHub Issue or related Pull Request, if one exists
N/A
## Description of change
I am developing [spice.nimabe.net](https://spice.nimabe.net), a static
frontend for SpiceAPI. It already supports generating and inserting new
card IDs, but SpiceAPI does not provide a way to discover cards already
configured on the game machine.
This makes the frontend inconvenient for existing users who generated
and saved their card IDs locally. They currently have to open each card
file on the game machine and manually copy its ID into the frontend.
This change adds a `card.get_cards` API function that enumerates the
current cards for readers supported by the running game. Each entry
contains:
- reader index
- canonical 16-character card ID
- `source`, set to `file` or `override`
- `file_name` for file-backed cards
File-backed cards expose only the save-file basename. Active `-card0`
and `-card1` values are returned with `source: "override"` and no
`file_name`, instead of representing an override as a file. spicefe
derives the useful default import names `card0` and `card1` from the
reader index. Reader enumeration uses `eamuse_get_game_keypads()` so
games with only one reader do not expose a second card.
This allows spicefe and other API clients to offer one-click import for
existing cards, without requiring users to manually find and copy their
card IDs.
Because card IDs are sensitive, `card.get_cards` can only be used when
the operator has configured an API password. No full filesystem path is
exposed. Missing, unreadable, and invalid cards are omitted.
The existing `card.insert` function remains unchanged. To avoid
requiring a password for the whole card module, this change adds support
for password requirements on individual API functions while preserving
existing module-wide password behavior.
Access to configured card paths and runtime card overrides is
synchronized because API requests may read them concurrently with
configuration updates.
The Python and Dart API wrappers and README documentation have been
updated for the new function and source metadata.
## Testing
- Ran the complete `src/spice2x/build_docker.sh` build successfully.
- Built all required 32-bit and 64-bit targets.
- Built the Windows XP-compatible 32-bit targets.
- Passed static import checks.
- Passed Windows 7 and Windows XP compatibility checks.
- Completed release packaging successfully.
- Verified both updated Dart wrappers are correctly formatted.
- Ran all 106 spicefe tests successfully.
- `git diff --check` passes.
*This implementation was prepared with assistance from OpenAI Codex.*
## Link to GitHub Issue or related Pull Request, if one exists
#0
## Description of change
Capturing a frame for the API stream made the game wait for
`GetRenderTargetData` in the middle of its present, roughly 1270us per
frame at 1080p. A 120Hz cab visibly lost frames for as long as a viewer
was connected.
The present thread now only issues a `StretchRect` into a render target
we own, which is queued rather than waited on, and a pool thread does
the readback and the pixel conversion. That takes the present thread
cost to 1-4us. Each snapshot is read on the request after the one that
took it, so the blit and its transfer have a full frame to land and the
read does not stall on the GPU either, at the cost of one frame of
stream latency.
Only streaming takes this path, and only on a device created with
`D3DCREATE_MULTITHREADED`. Screenshots, `capture.get_jpg` and the
`THREAD_BAN` models keep the existing inline readback unchanged.
Also raises the x264 encoder from `i_threads = 1` to 4, which was
holding a 1080p60 stream to 41fps and making a keyframe cost 12.7ms
against 6.6ms for an ordinary frame. Capped rather than automatic
because this encodes on the same machine it is capturing.
## Testing
tested against iidx33, which was the most sensitive to frame drops
## Link to GitHub Issue or related Pull Request, if one exists
#0
## Description of change
MSE method of streaming video needs the dimensions up front. This new
api delivers that. Also provides an easier way to determine which
screens are available and a way to discover port number.
## Testing
See pending changes in substream project
## Link to GitHub Issue or related Pull Request, if one exists
#0
## Description of change
Moves the API capture readback off the game's Present thread while a
video stream client is connected.
The readback is a `LockRect` plus a memcpy of the whole back buffer,
roughly 635us at 720p and 1270us at 1080p. On the Present thread that
comes out of the game's frame budget: TDJ (at 120Hz) dropped to 117fps
with a 60fps stream running, and reading on a pool thread instead gave
the full 120 back.
Only streaming takes the off-thread path, gated on a new
`capture_pump::screen_claimed()`.
Screenshots, one-off API captures, and `THREAD_BAN` games all keep the
existing inline read for compat reasons. A pool thread in `LockRect`
while the Present thread sat inside `GetRenderTargetData` deadlocks DDR
X2 for example.
`CLAIMED[]` becomes `std::atomic<bool>` so the capture path does not
take a lock on the Present thread. The read pool has a single worker so
frames cannot be enqueued out of order, and both capture pools are never
destroyed so a late read cannot queue onto a torn-down pool.
The capture pipeline itself is unchanged: `GetRenderTargetData` is still
synchronous on the Present thread.
## Testing
DDR X2
World
IIDX TDJ
SDVX VM
## Link to GitHub Issue or related Pull Request, if one exists
n/a
## Description of change
Adds `Access-Control-Allow-Origin: *` to the video stream server's
responses.
## Testing
Tested with substream project.
## Link to GitHub Issue or related Pull Request, if one exists
fixes#875
## Description of change
Adds `-apistream`, an optional HTTP video stream of the mirrored screen.
It listens on the API port +2.
Two endpoints, sharing the same `screen`, `fps` and `q` parameters:
/stream.mjpg JPEG frames, for clients with no container support
/stream.h264 H.264 annex-b, for an app driving MediaCodec or
VideoToolbox itself
One encoder per connection, fed by a per-screen pump that always hands
over the newest frame, so a slow reader drops frames instead of building
a backlog. `capture.get_jpg` behaviour is unchanged.
Additional documentation for developers:
https://github.com/spice2x/spice2x.github.io/wiki/Video-Stream
## Testing
## Link to GitHub Issue or related Pull Request, if one exists
#0
## Description of change
Significantly speeds up API screen capture and D3D9 screenshots saving.
Two reasons for doing this:
1. We now have a 4K game (GITADORA) and existing capture code was taking
multiple seconds.
2. Renewed user interest on streaming as we have a couple more companion
apps in active development.
**API screen capture (streaming), 1280x720:** 14.3ms -> 6.3ms per frame.
Back buffer copies go to pooled `D3DPOOL_SYSTEMMEM` surfaces via
`GetRenderTargetData` instead of allocating a lockable render target
every frame, and TooJpeg is replaced with libjpeg-turbo (encode 9.8ms ->
3.0ms). MSAA remains unsupported
**Screenshots for GITADORA arena model, across 4 screens with one of
them 4K**: 4068ms -> 124ms. `D3DXSaveSurfaceToFileA` is replaced with
fpng (encode 4043ms -> 76ms) and the screens encode in parallel.
Dropping D3DX also removes the `d3dx9_43.dll` ... `d3dx9_24.dll` probing
loop, so screenshots no longer fail outright on machines with no D3DX9
runtime installed.
Screenshot surfaces are read on the present thread, so no D3D call
reaches another thread for screenshots. This fixes a hang in DDR X2
introduced earlier in the branch: its device has no internal locking,
and reading the surface on a pool thread while the present thread sat
inside `GetRenderTargetData` left the game's own render thread
deadlocked.
## Testing
- **GITADORA** (arena model, D3D9Ex, 4K main plus three subscreens,
windowed) with
`-screenshotsub`: three sets of four screenshots, images verified
correct. Completion
order differs between sets, so the screens really are encoding in
parallel.
- **LovePlus** (KLP, plain D3D9, 768x1360): covers the inline path used
by games whose
image processing must not leave the present thread.
- **API screen capture** through a companion app: live video correct
throughout.
- **Print Screen** bound as the screenshot key: the clipboard copy
succeeded on every shot.
- Quitting the game after capturing leaves no `IDirect3DDevice9`
reference count warning,
so the pooled readback surfaces are released along with the device.
## Link to GitHub Issue or related Pull Request, if one exists
Related: #859 and #862.
## Description of change
Three SDVX (Valkyrie model) fixes:
* **API touch went to the wrong window.** SDVX registers touch on both
its Main Screen and Sub Screen windows, and the native injector kept
whichever attached last. The touch surface is now published explicitly:
sub screen window when windowed, main window in fullscreen.
* **Landscape never rotated API coordinates.** Synthetic contacts bypass
`transform::hardware_to_game`, so they missed the rotation a real finger
gets. Applied before injection now, as portrait already did. Extracted
to `sdvx_landscape_rotate` so both paths share it; gated on the native
path since `wintouchemu` rotates via the subscreen overlay instead.
* **`-sdvxnosub` didn't hide the sub window in fullscreen.**
`ShowWindow_hook` had branches for GITADORA, pop'n and IIDX but not
SDVX. Added the missing one.
## Testing
Tested Nabla
## Link to GitHub Issue or related Pull Request, if one exists
#0
## Description of change
Fix mouse and touch API handling for gitadora.
* hook native touch in every arena window mode (previously only using
wintouchemu for 1 windowed mode)
* publish the touch window (SMALL panel, or the main window drawing the
overlay) and only attach injection there
* route touch on the SMALL window straight through instead of the
overlay transform, and accept the mouse only there
* fix spiceapi by forcing 800x1280 touch canvas size
* run API capture before the subscreen present to fix api not showing
sub image
## Testing
Test:
gitadora 1/2/4 windowed mode, fullscreen
iidx /sdvx / popn windowed/fullscreen for regressions
## Link to GitHub Issue or related Pull Request, if one exists
Fixes#858
## Description of change
Fix orientation being wrong in windowed mode when API touch is used
Fix mouse no longer working when API client is in use.
This PR only affects windowed mode SDVX.
## Testing
## Link to GitHub Issue or related Pull Request, if one exists
Fixes#858
## Description of change
Landscape mode needs another adjustment applied to touch coordinates.
This fixex both native touch and wintouchemu paths.
## Testing
Tested landscape, also with full screen resolution changes.
## Link to GitHub Issue or related Pull Request, if one exists
Fixes#852
## Description of change
Add SDVX valk cab tape LED output over API
## Testing
tested with custom python script over api
## Link to GitHub Issue or related Pull Request, if one exists
n/a
## Description of change
- Replace unnecessary precision timers with standard sleeps to reduce
wakeups and background CPU usage.
- Reuse buffers in raw input, touchscreen, and HID output paths to
eliminate steady-state allocations.
- Pre-index HID button groups and correctly process batched HID reports.
No functional changes.
## Testing
## Link to GitHub Issue or related Pull Request, if one exists
Fixes#814
## Description of change
Touches via the API wasn't being delivered to native touch hook;
implement this, which closes the last gap we have vs. wintouchemu.
While we're here, implement popn subscreen support for companion as
well.
## Testing
Tested for iidx/sdvx/popn
Add a timeout and skip signalling to the D3D9 capture path so the API
thread no longer waits indefinitely when Present stops during loading.
Return the last successful JPEG frame as a fallback.
## Link to GitHub Issue or related Pull Request, if one exists
#746
## Description of change
* Fix `capture.get_jpg` blocking indefinitely when D3D9 screen capture
cannot
complete during game loading (e.g. no Present calls).
* Add a 2-second timeout to `graphics_capture_receive_jpeg()` and cancel
pending capture requests on timeout.
* Signal capture skip from D3D9 failure paths via
`graphics_capture_skip()`.
* Cache the last successful JPEG per screen and return it as a fallback
when a new capture fails.
Tested with **stock iOS Spice Companion** only. I have not tested
Android or
other Companion clients. The iOS client implementation differs from the
others; this fix is server-side only.
No config file, CLI option, or API schema changes.
## Testing
* Built locally with llvm-mingw cross-compiler (WSL).
* Docker build (`src/spice2x/build_docker.sh`): Pending
* Tested with stock iOS Spice Companion on iPad: Mirror remains
connected
during game loading instead of returning to KeyPad.
## Link to GitHub Issue or related Pull Request, if one exists
n/a
## Description of change
Adds toast notifications to the overlay.
Toast notifications are transient windows (shown for about 4 seconds).
These are drawn on top of the game even when the overlay is inactive
(i.e., when no window is open), and does not cause any input sink
behavior.
Add toasts for some common user-driven actions that could use a
notification, such as:
* screenshot
* card insert
* PIN macro
* API client connect/disconnect
* virtual printer
* screen resize toggle / scene switch
Add an option to change where the toasts are displayed, or to turn it
off entirely.
Add SDK function for it, and add it to the samples.
## Testing
Tested DDR (shown at bottom right, as is for most games) and RB (shown
top right by default).
## Link to GitHub Issue or related Pull Request, if one exists
Fixes#681
## Description of change
`Sleep` and `sleep_for()` can be very inaccurate and varies depending on
what the OS gives us...
### `timeBeginPeriod(1)`
On boot, we are now calling `timeBeginPeriod(1)`, which affects the
whole process but makes `Sleep` more accurate. There is some risk here
if any game was relying on doing things like `Sleep(1)` and expecting it
to run for 15.6ms. Most games already call `timeBeginPeriod(1)` in the
game engine, though not the whole time, so I'm hoping that this is not
too impactful.
### Opt out of Win11 power throttling
Ensure that timer resolution change above is respected even when the
window is occluded / minimized by opting out of throttling via
`PROCESS_POWER_THROTTLING_IGNORE_TIMER_RESOLUTION`.
### Use Win10 high resolution timer instead of Sleep
On Win10 1803 and above, there is a new OS-level API for high resolution
timers; if this is available, use it
(`CREATE_WAITABLE_TIMER_HIGH_RESOLUTION`). Worth noting that WINE
doesn't support this currently.
If not, fall back to `Sleep`, which is significantly better than
`sleep_for()` in my experiments.
Callers of Sleep / sleep_for were replaced with this new timer. Most of
them anyway; calls to Sleep() with more than 100ms+ was left alone.
### Add an option as a chicken bit
To opt out I'm adding a new option called `Use Legacy Timers` which will
revert to behavior before this PR. The code paths that switched from
`sleep_for` to `Sleep` will remain in place though, not affected by the
option.
## Expected changes
In some I/O emulation modules, poll threads may run more frequently,
resulting in lower latency.
It also means that spice overall may use more CPU resources and power.
If you don't like this, you can always enable the option to opt out;
e.g., if you're on old arcade cab PC.
## Testing
DDR p4io - ok
drs touch hook - ok
IIDX camera hook - ok
CCJ trackball - ok
## Link to GitHub Issue or related Pull Request, if one exists
#676#452
## Description of change
When spice SDK or API sets or clears button and analog override, we
don't tell MDXF about it, so it may fall behind. This is a special case
for DDR MDXF I/O emulation. For Naive/Xinput I think we are getting
lucky because they usually also generate WM_INPUT events.
Update the C++ SDK sample to provide more interesting example of
emulating DDR arrows with the keyboard.
## Testing
As noted in #567, a filesystem path that contains non-ascii will break a
lot if using a clang toolchain.
Luckily, fmtlib has a lossy utf8 convert when you use it to print a path
(after including `fmt/std.h`). The vast majority of this diff is just
removing `.string()` from paths inside loggings calls.
There are some callsites I _didn't_ touch, mainly the options, because
it would be an ABI break to change those to be wide strings and I cbf
looking into settings upgrades. There are also some spots (avs mountpath
remapping, for example) where the path is guaranteed to be ascii, so I
didn't modify them.
ImGui doesn't appear to easily support wide strings (I mean, surely it
does, but I'm not gonna look too far into it) so I mostly just left
those alone too, with a few spots modified to re-use fmtlib's lossy
utf8.
Some of the changes are basically never gonna be hit IRL, like who would
put a file with a non-ascii _extension_ along with their modules? But
the diff is (I hope) pretty easy to validate as OK.
Testing has been somewhat minimal, I fired up the GCC build of spice2x
in a dodgy folder name, got mojibake (running via wine in linux so take
that as you will), ran the unmodified clang spice and crashed the same
way the reporter did. After modification, I get the exact same mojibake
so I assume if the terminal enjoys utf8 it'll display OK.
Claude (only used for review) thinks the commit is fine but is annoyed
that I use `fmt::detail` in the appdata censoring, which is part of the
private API; personally I don't care because it's pretty stable.
## Link to GitHub Issue, if one exists
N/A
## Description of change
Changes the array size used for collecting touch inputs from the API to
be dynamic according to the size of the input data, rather than fixed.
This fixes a crash.
## Testing
An API client was made sending touch states, both valid and invalid
(oversized). The valid states are handled identically in the test menu
while the oversized states no longer crash.
## Link to GitHub Issue, if one exists
#41 (partially)
## Description of change
This PR adds support for outputting IIDX tape LED data via spiceapi
(`iidx_tapeled_get()`). It returns the data as a `dict`, with the key
being the LED name and the value being the list of RGB values. You can
optionally pass in a variadic list of names to filter the returned data.
The new API function is only implemented in Python since I'm not able to
test the wrappers in other languages.
An example script using the new API is available in this Beef Board PR:
https://github.com/HWXLR8/beef-board/pull/133.
Demo use case: https://streamable.com/pq5e1e
## Testing
Tested in IIDX 32 with custom Beef Board firmware along with a Python
script to read and send the tape LED data over USB.
## Link to GitHub Issue, if one exists
n/a
## Description of change
Allow the server to override client's quality options for screen
mirroring.
On Windows / Android versions (flutter-based) clients, this is already
in the options of those apps. However, the iOS version lacks the quality
options.
## Testing
Tested with iOS and Windows clients.
## Link to GitHub Issue, if one exists
Fixes https://github.com/spice2x/spice2x.github.io/issues/179
## Description of change
Previously, lights.read() did not accept any parameters, instead it
returned all lights implemented by the current game.
Problem is that this is too large for some embedded implementations;
also just inefficient.
Add code to accept light names as strings and only return those in the
result.
For performance, use `robin_hood::unordered_map` to speed up the lookup
operation, which was previously a linear search.
Add this to Python wrapper library. Did not bother with others.
Also, as a bonus - update lights.write_reset() to also accept a flat
list of strings, as opposed to array containing a string, which was
silly.
## Compiling
➕
## Testing
Tested using spiceremote and writing a short Python program.
## Link to GitHub Issue, if one exists
Fixes#263
## Description of change
Add "scenes" to screen resize.
* UI changes - F11 menu now has a "scene switcher" for `Scene 1/2/3/4`.
Clean up other UI bits.
* Add hotkeys for `Screen Resize Scene 1`, `Screen Resize 2`, and so on.
* Detect hot key and trigger resizes.
* Add additional objects to screen_resize.json for saving and loading
the new scene settings.
* Existing screen resize data is backwards compatible (can be
loaded/saved to JSON), and is treated as `Scene 1`.
* Add API endpoint for `Resize`, allow toggling resize on/off, switching
active scene
* Add Python wrapper. Did not bother with the other wrappers (Dart, C++,
etc)...
## Compiling
🥇
## Testing
Still testing, but seems to work fine for TDJ.
bicarus