- Remove the MOCKTAIL_VR_TEST_SESSION_LOSS_AFTER fault injection; nothing
used it.
- The scripted head pose and the eye evidence dumper only work in builds
configured with -DMOCKTAIL_VR_DIAGNOSTICS=ON (default OFF); release
builds ignore their environment variables.
- Remove FindFmodInputCaptureProfile and HasExpectedFmodInputCaptureVtable
(only their own tests called them) and the never-defined
PrepareCaptureLocked declaration; the audio bridge files match upstream
again.
- docs/VR.md lists every MOCKTAIL_VR_* variable with its default, linked
from the README.
- MOCKTAIL_VR_HAND_POSE_SOURCE=grip now rotates hand records from the grip
pose. base_ori started as the identity quaternion, so the norm check
never fell back to it and hand rotation was dropped.
- Hand records cancel the engine's stored hand pitch (read from
hand_pitch_storage_rva by the pointer hook, 20 until read) instead of a
hard-coded 20, so they agree with the pointer ray.
- Shutdown never rewrites the aim-pointer prologue while guest threads may
be executing it; the jump stays and only forwards once the bridge is gone,
and a later activation reuses it.
- Guest hooks count themselves in flight before loading the bridge pointer;
Shutdown unpublishes the bridge and waits for in-flight calls to drain,
so no hook keeps a pointer to a destroyed bridge.
- Roblox's original state and eye getters are kept for the process
lifetime. A hook reached with no bridge, including after a failed vtable
restore, forwards to them instead of returning an unfilled state buffer.
Restoring the vtable no longer depends on the pointer hook.
- The activation log names the vtable slots in the right order.
- Once the guest eye format is known, the eye swapchains are recreated
once in the format that keeps its bytes: the eye format, else its
UNORM/SRGB sibling (raw copy), else the other channel order with the same
sRGB-ness (a blit that only swizzles). Mixed order-and-gamma cases no
longer go through a gamma-converting blit; if only that remains, it is
logged.
- The desktop mirror segment is created with O_EXCL after removing our own
stale segment (a crashed run's leftover), and must be private to this
user before eye images are written to it.
- A new frame_mutex_ is held for the whole XR frame cycle and by every path
that can destroy the session (taken before mutex_). That lets the cycle
release mutex_ while it blocks in xrWaitFrame, so guest render-thread
hooks no longer wait out the runtime's pacing. Swapchain-image and fence
waits stay under mutex_; they return promptly unless the GPU or
compositor stalls, and they run while guest eye bindings are copied.
- Every frame error, including the swapchain-image and copy-fence
timeouts, now tears the session down recoverably instead of ending XR
output for the rest of the run.
- Session recovery backs off (1, 2, 4, 8 s) and recreates the XrInstance
after an instance-loss event, an instance/runtime failure, or three
failed attempts, so VR returns after SteamVR or WiVRn restarts instead
of retrying forever.
- While no XR session runs (headset idle, session not READY, recovering)
presents are paced at 60 Hz outside the backend lock instead of running
unthrottled.
- A PrepareHostPresent failure that consumed nothing (copy resources
unavailable, failed fence reset or submit) presents normally instead of
failing the guest's vkQueuePresentKHR; only device loss is propagated.
- A runtime/guest GPU mismatch falls back to the plain host device and
keeps the game running without headset output, naming both GPUs.
- A failed EGL attach on the GLES path no longer makes window init fail;
the game runs without headset output and the bind is not retried.
Queue and device-idle locks in the Vulkan adapter are now taken only after
an armed OpenXR backend calls mocktail_vulkan_enable_queue_sharing(). The
guest synchronizes its own queue use; only the runtime and the eye copies
add a second user. Without VR, and in builds without it, submits and
presents take no locks.
FFlagDebugEnableVREmulator is only written when VR is active, so a user's
own fflags.json value is no longer overridden on every launch.
Config loading is back after the single-instance lock and the support
bundle guard. A non-VR build still rejects --vr before creating any data
or state directories, from the command line alone; a vr.enabled config in
such a build is rejected once the config is loaded.
An explicit --vr on a Roblox build without VR support now exits with the
build's version and Build ID, the builds VR supports, and how to select one
(--roblox-lib). When VR is only enabled in the config, the launch continues
without VR and logs the same details. The bridge's own Install error names
both as well. The verified build contracts carry their version labels and
move to namespace scope so they can be listed.
VR now stays off at runtime in every build until --vr or vr.enabled: true,
so a normal launch on a VR-capable build never tries to reach an OpenXR
runtime and no longer exits when none is running.
MOCKTAIL_ENABLE_VR defaults to AUTO: OpenXR support is built when the
third_party/OpenXR-SDK submodule is present and skipped with a notice
otherwise, so source packages that do not fetch submodules (Nix, AUR) keep
building. ON still requires the submodule. mocktail-vr-probe is only built
and installed in VR builds.
CI gains an x86_64 job that builds and tests with MOCKTAIL_ENABLE_VR=OFF.
The config docs and README describe the new defaults.
Mark the 3092 profile experimental. It still runs when selected with
--roblox-lib, but it no longer becomes the preferred update target for
every x86_64 user, so the shipped payload metadata (2998) is again the
preferred profile.
Select the 2998 entry by Build ID in the partial-input profile test instead
of relying on manifest order, and drop the shipped 3092 entry before the
test re-identifies 2998 as 3092.
Align the example config's VR build wording with the bootstrap template.
When Roblox asks for a keyboard (showKeyboard), a world-locked panel
opens below eye level as an OpenXR quad layer. Controllers point at keys
and the trigger types; text and key presses go through the same platform
events as the desktop keyboard, so the focused TextBox edits identically.
The layout is US QWERTY without the function row or Ctrl/Alt, with one-shot
Shift, Caps Lock, Tab, Backspace auto-repeat, Enter and Close. A trigger
aimed at the panel is withheld from Roblox so a key press is not also a
click. A text field above the keys mirrors the focused TextBox, whose
desktop overlay never reaches the eye images. MOCKTAIL_VR_KEYBOARD=0 turns
it off.
vr.frame_rate_limit / MOCKTAIL_VR_FRAME_RATE_LIMIT replaces
graphics.frame_rate_limit while VR is active and takes the same values.
It defaults to unlimited so xrWaitFrame paces Roblox at the headset
refresh rate instead of Roblox's own 60 FPS cap.
MOCKTAIL_VR_RENDER_SCALE scales the runtime's recommended eye extent.
SteamVR's recommendation for a Quest 2 at 90 Hz (2244x2352 per eye) missed
the 11.1 ms budget and halved the rate to 45 FPS.
MOCKTAIL_VR_UNPACED=1 stops guest frames from blocking in xrWaitFrame:
the open XR frame spans several guest frames and is submitted once per
display period, for measuring maximum render rate.
Roblox maps some rbxasset:// names to downloads and opens no local file for
them. For each file under content/ in the overlay folder, change its name in
the loaded libroblox so Roblox opens the local file.
Idea from @dim-ghub in #20.
Roblox 2.742 moves the process-load constructor behind the JNI singleton
initializer and renames registers in the registry initializer. The deriver
finds each leading constructor by its code, and the loader checks that host
bridges run the checkpoint and JNI singleton constructors and skip the two
wiring entries.
strerror_r: Bionic's message table covers errno 0..133 except the
unassigned 41 and 58. Numbers outside that set now get "Unknown error N"
without consulting the host, so the text is the same on glibc and musl and
no longer relies on comparing against strerror_r(-1).
atoi: glibc's atoi is (int)strtol() like Bionic's, so the host symbol is
used there as before. The Bionic shim is registered only on libcs where
that is not true (musl). linker_registry_test expects the matching result.
Brings in 3e6f63e (drop libexecinfo), aa31cdf (Chimera/musl portability),
bc10810 (Bionic atoi, canary signal exit codes) and cb2e9dd (CI fixes).
Conflicts:
- tests/native_update_test.cc: keep both <cerrno> and <csignal>.
- tests/webview_helper_launcher_test.cc: keep the existing python3 socket
read; it already solves the dash multi-digit fd redirection problem that
upstream fixed by switching the helper to bash.
A driver can reject a present mode the device did not enable. Retry once with the mode the game asked for, except after errors another mode cannot fix.
Fill the 65 unimplemented JNI table slots with stubs that log once and return zero, so a call no longer jumps to address 0. A release of a handle with no reference count now returns, so a double release cannot queue one slot twice and give two objects the same handle.
vii