With DARLING_METAL_MAC_GPU_NAME=1, MTLDevice.name is the GPU the
MacBookPro16,1 that IOKit reports shipped with, from the host's vendor:
"AMD Radeon Pro 5500M" (vendor 0x1002) or "Intel UHD Graphics 630"
(0x8086). NVIDIA keeps its own name so the game never takes another
vendor's driver paths. recommendedMaxWorkingSetSize and every other value
stay the host's, which was the risky part of #61's identity.
The Riot launcher turns it on by default (0 or empty turns it off); other
Darling apps keep the host name. DARLING_METAL_GPU_NAME still wins.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
system_profiler now answers SPHardwareDataType and SPSoftwareDataType (and
lists them in -listDataTypes). Model, serial number, platform UUID and
provisioning UDID come from IOKit's IOPlatformExpertDevice via ioreg; CPU,
cache and memory from sysctl; the OS version from sw_vers and uname. A
field the guest does not provide is left out.
#61 instead read a separate darling-smbios.conf written by a new
generator (gen-darling-smbios.py), with hardcoded fallbacks (an Intel Core
i9, macOS 14.7, SIP enabled). iokitd reads darling-platform.plist, so
system_profiler and IOKit could report different serials and models on the
same prefix, and the generator overwrote identities imported with
configure-darling-platform.py. Reading the guest's own APIs keeps a single
source of truth: configure-darling-platform.py remains the only way to set
or provision an identity, and it is still never run by install or launch.
tests/test_system_profiler.py runs the script under sh with stub guest
commands and runs in CI.
Split from #61.
Co-authored-by: DevWebeloper <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From #61, without the default "AMD Radeon Pro 5500M" / 4 GB identity: the
host GPU's name and Vulkan working set stay the default for every app, and
the existing DARLING_METAL_GPU_NAME and DARLING_METAL_VRAM_MB overrides
remain the opt-in. Reporting 4 GB on a 2 GB card invites allocation
failures, and under-reports larger cards.
Kept from #61:
- registryID returns MTLDeviceInfo::kDefaultRegistryID instead of 0; a
real Mac never reports 0 and apps key devices by it.
- Vertex buffer layout and attribute arrays release the old item after
retiring it; the retire step read the item after the release could free it.
- newDepthStencilState, newSamplerState and the device name return nil (or
the fallback name) when Indium throws, instead of unwinding through
Objective-C.
- MTLDepthStencilGetIndium is declared only with DARLING_METAL_ENABLED.
- The replaceRegion trace probes at most `width` bytes, which stays inside
the first row for every pixel format.
metal.patch is regenerated in its existing file order, so the diff is
these changes only. tests/test_metal_device_info.py compiles the host
check against MTLDeviceInfo.h as shipped in the patch and runs in CI.
Split from #61.
Co-authored-by: DevWebeloper <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
After the clients quit, the session kept Darling running: Riot and League
crash handlers never exit under Darling, and the idle check counted them as
clients. Ctrl+C or closing the terminal killed bubblewrap, whose
--die-with-parent then killed nix-lifecycle.py before it could shut Darling
down, which left a live server marker behind.
- nix-idle.py ignores crash handlers when deciding whether a client still
uses the prefix.
- The launcher starts bubblewrap with INT, QUIT, HUP and TERM ignored, and
env restores them inside. The supervisor turns them into a stop request.
- nix-lifecycle.py polls for a signal or stop request and then shuts Darling
down without waiting for clients. It stops retrying a failed shutdown once
the server is gone, and waits for or kills its own command afterwards.
- darling-lol --stop makes the stop request and waits for the session. If
the session does not exit within 30 seconds, or no session holds the lock,
it kills the prefix's darlingserver and launchd like darling shutdown,
after checking that the marker PID is this user's darlingserver for this
prefix.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
RCLOL_FILTER_HOST_ENV=1 makes exec-guest.py drop host session metadata
(WAYLAND_DISPLAY, XDG_SESSION_*, XDG_CURRENT_DESKTOP, GNOME/KDE/GDM,
systemd, SSH agent, terminal, toolkit, Flatpak/Snap/AppImage, exported
shell functions) and set SHELL to the guest's /bin/zsh. It is off by
default; RCLOL_KEEP_HOST_ENV lists names to keep anyway.
Unlike the denylist in #61, the filter runs after the launcher sets its
GPU tuning, keeps everything Darling and the host graphics and audio
libraries read (DISPLAY, XAUTHORITY, XDG_RUNTIME_DIR and the XDG base
directories, D-Bus, PulseAudio/PipeWire/ALSA, DRI_PRIME, MESA_*, __GL_*,
__NV_*, VK_*, LD_PRELOAD for gamemoderun, LD_LIBRARY_PATH), and lives in
one place instead of three.
tests/test_exec_guest.py runs in CI; the launcher check also verifies
that MESA_NO_ERROR=1 and __GL_THREADED_OPTIMIZATIONS=0 reach the guest
with the filter on.
Split from #61.
Co-authored-by: DevWebeloper <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A terminal, IDE or service manager can leave descriptors open in the
launcher, and the guest inherited all of them. tools/exec-guest.py closes
every descriptor above stderr and then execs the guest command.
A shell loop over /proc/self/fd cannot do this portably: dash only parses
single-digit descriptors in redirections, so "exec 10>&-" runs a command
named 10 and ends the launcher on Debian and Ubuntu.
The launcher offline check now leaks descriptors 3, 4, 12 and 200 into the
launcher and asserts the guest sees only 0, 1 and 2. It runs in CI.
Split from #61.
Co-authored-by: DevWebeloper <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Chromium on macOS takes its temp directory from NSTemporaryDirectory(), which is
unusable under Darling. Electron opens files in app.asar asynchronously by
copying them to a temp file, and reports the failed copy as "not found", so the
Riot client showed a "Not Found" page instead of its UI. Running the client
binary as Node reproduces it: fs.readFileSync on app.asar/dist/renderer/index.html
works, fs.createReadStream fails with ENOENT, and MAC_CHROMIUM_TMPDIR fixes it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
With __GL_THREADED_OPTIMIZATIONS=1, NVIDIA 580 (the last branch for Pascal
GPUs) with egl-x11 1.0.6 rejects the X11 EGLDisplay with EGL_BAD_DISPLAY on
every thread that has a context current. A native EGL program reproduces it
outside Darling. In the Riot client the main thread loses EGL after its first
window frame, so CGLCreateContext fails with error 10004 and the game window
stays black. Default the launcher to 0; setting the variable restores 1.
CGLDestroyWindow also returns early for the NULL window CGLGetWindow gives
when eglCreateWindowSurface fails, instead of passing EGL_NO_SURFACE on.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Keep the merged per-device completion budget, bounded parsing, idle blocking and timeout accounting. Scope the registration/blocking handshake to each device, restrict live batch windows to the timer resolution, and test registration interleavings, spin-disabled polling, shutdown, opt-out and independent devices with actual patched methods.
Co-authored-by: SypherrDev <[email protected]>
Ignore both invalidation methods on the shared session; deliver task completion through the owning session delegate queue. Extend the standalone probe to exercise singleton reuse and a suspended custom queue. Require stable old-client process identities and complete window data before optional recovery, and continue launch after shutdown failures.
Rebuild the remaining NVIDIA changes on current main, preserving the merged visual, draw-buffer, descriptor ownership and mixed Metal-tree fixes. Scope native-thread forwarding to resize notifications and make completion polling per device, bounded and part of the caller timeout budget.
Co-authored-by: Pirrandi <[email protected]>
Security reads the system trust domain from
/System/Library/Keychains/SystemRootCertificates.keychain and
SystemTrustSettings.plist (TrustSettingsSchema.h). Darling ships neither, so
SecTrustSettingsCopyCertificates(kSecTrustSettingsDomainSystem) returns
errSecNoTrustSettings (-25263); Chromium's known_roots_mac and Riot's
CertBundle then report that they cannot read the system roots.
ensure-darling-root-certificates.py parses each host root with a minimal DER
walker and writes the trust record keyed by the uppercase SHA-1, with the
encoded issuer Name, the serial INTEGER's content octets, modDate and an
empty (unrestricted) trustSettings array. The keychain is built with the
guest security tool in one darling shell, in a private temporary directory,
and published from inside the guest so a running Darling sees it, falling
back to the host when the guest cannot write the directory. A digest stamp
skips later runs without starting Darling. The launcher warns and continues
on failure; install-runtime.sh runs it before stopping Darling.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
-[MTLDevice name] returned "Darling Indium Device" for every GPU. It now
returns the selected Vulkan device's name in macOS Metal style: the
trailing driver note in brackets and the (R)/(TM) marks are removed, and
NVIDIA names are unchanged. The name is never replaced with another
vendor's. recommendedMaxWorkingSetSize is the device-local heap size on
discrete GPUs and all heaps on integrated/CPU devices, with the old 1 GiB
as fallback. Both are computed once per device.
hasUnifiedMemory stays YES by default: synchronizeResource:/
synchronizeTexture: are missing and Private buffers created with data
abort in Indium, so reporting NO on discrete GPUs would send apps down
unimplemented paths. DARLING_METAL_HOST_UNIFIED_MEMORY=1 opts in to the
GPU-type answer. DARLING_METAL_GPU_NAME and DARLING_METAL_VRAM_MB
override the name and working set.
The cleaning logic lives in private-include/Metal/MTLDeviceInfo.h and is
covered by the host check tools/test-metal-device-info.cpp. Metal now
links indium_private for the Vulkan properties, and the runtime build
script adds Indium's private headers before compiling MTLDevice.mm.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
League enters relative mouse mode on GNOME (Xwayland) and inside gamescope, but
keeps a visible cursor and warps it every frame at its own mouse speed. Xwayland
cannot move a visible pointer, so those warps never took effect: the virtual
cursor, where clicks land, drifted hundreds of pixels from the cursor on screen
and never reached the screen edges, which also stopped edge panning.
Where warps are not honoured, the virtual cursor now follows the physical
pointer and the game's warps leave it in place. With a hidden cursor, or on a
server that honours warps, nothing changes. The host model test gains a case
replaying the drift from an in-game trace.
Tested manually in the practice tool under gamescope 3.16.31 on GNOME 51 with
AMD RADV: clicks land on the cursor and edge panning works.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Port the fullscreen policy omitted when adapting PR #12. Track requested fullscreen state across frame and style changes, exclude unmanaged windows, and use EWMH REMOVE when leaving fullscreen.
Extend the native geometry probe and document passive tracing for the reported vertical click offset, whose cause remains unconfirmed.
CGDisplayScreenSize returned 0x0 for any display with EDID (upstream left
that branch unimplemented). League divides by it for its camera aspect
ratio, so the view-projection x row became NaN: every 3D vertex was
discarded, the scene showed only its blue clear colour, and world clicks
were ignored. Read the size from the EDID detailed timing (mm) or basic
parameters (cm), fall back to 96 dpi, and use the main display for an
unknown id.
CGDisplayCopyAllDisplayModes returned NULL for display 0 (a window off
every screen), which League passes straight to CFArrayGetCount. Fall back
to the main display, then to an empty array.
Add riot-display-modes-check covering both for the main display,
display 0 and an unknown id.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Adapt the geometry fix from PR #12 to the current relative-mouse implementation. Keep display IDs independent of primary-first screen ordering, translate window and pointer coordinates consistently, and invalidate geometry on both RandR event types.
Handle failed RandR snapshots and free EDID data. Add isolated native geometry regression coverage and use the configured SDK when linking the X11 probe.
Validated all 30 component patches, the X11 and CoreGraphics builds, six native geometry groups, and the existing relative-mouse checks. The user confirmed the installed fix works on 2026-10-16.
Co-authored-by: Maka <[email protected]>
CGAssociateMouseAndMouseCursorPosition(false) used to warp the pointer to the
window center, freeze the reported position there and treat the game's own
CGWarpMouseCursorPosition calls as motion. On Xwayland, which ignores warps
while the cursor is visible, this left League's cursor stuck in the middle.
The backend now keeps a frozen virtual cursor: physical motion only produces
deltas (deltaY positive downwards), application warps move the virtual cursor
and their echoes are dropped, mouseLocation returns it, and the pointer is only
warped back where the server honours warps. The grab no longer uses the
invalid FocusChangeMask and is released and retaken across focus changes.
setCursor: no longer re-sends XDefineCursor/XFlush when the window already has
that cursor; League set it about 190 times per second.
Also carries the opt-in input trace (/tmp/rclol-input-trace in the guest).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Traverse the layer tree before compositing so nested Metal layers consume queued drawables and recycle their pool. Add a nested presentation probe covering pool reuse and Vulkan-to-OpenGL pixel readback.
NVIDIA's GL rejects GL_LINEAR_TILING_EXT with GL_INVALID_VALUE, so the
linear internal image could not be imported and the window stayed black.
Pick the tiling by GL_VENDOR; other drivers keep linear.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Wait up to 2 s for the cancellation callback instead of one 50 ms run-loop
pass, count only callbacks without stored addresses so a completed lookup
cannot pass as a cancellation, print the failing stage, and have the runner
require the exact success line since Darling can exit zero after a trap.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
League sets framebufferOnly=NO on its CAMetalLayer, and QuartzCore read
that as needing host-visible coherent memory, so the game's final
render target and the linear image handed to OpenGL both lived in
system RAM (type 2 on the RX 6600) and crossed PCIe every frame.
framebufferOnly=NO only lets Metal sample or blit the texture: nothing
maps either image (replaceRegion aborts, getBytes is a stub, Indium
maps only Shared/Managed buffers).
Drawables now take the first allowed DEVICE_LOCAL type whatever
framebufferOnly says; the guest file /tmp/metal-no-drawable-vram
restores the old placement. metal-device.log shows drawable-vram=on and
storage=vram.
A native EGL/RADV test exported the drawable's linear image from VRAM
and imported it into GL with no errors and identical pixels. In a
practice-tool baron fight on the RX 6600 with the same champion it ran
about 20 FPS faster and the same when idle; not yet measured in an ARAM.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Indium and QuartzCore took the first memory type a resource allowed. The
Vulkan spec orders a type with no flags before every DEVICE_LOCAL type,
and NVIDIA exposes such system-RAM types, so on NVIDIA private buffers,
textures and framebufferOnly drawables could land in system RAM. They
now take the first allowed DEVICE_LOCAL type and fall back to the old
choice; shared and managed resources are unchanged.
The drawable's exported image is a dedicated Vulkan allocation, but the
OpenGL import never set GL_DEDICATED_MEMORY_OBJECT_EXT, which
EXT_external_objects requires. Mesa ignores it; it is the suspected
cause of the black screen on NVIDIA in issue #3. The import's GL errors
were never checked (reportGLErrors is compiled out), so a failure was
silent.
/tmp/metal-device.log now lists the selected device's memory heaps and
types, the memory type chosen per resource kind (with the old choice and
changed=yes/no), and the result of the first drawable and semaphore
imports into OpenGL (later ones only on failure).
Practice match on the RX 6600: all choices changed=no, import
dedicated=yes with no GL errors, semaphore ok, match=yes. Not run on
NVIDIA.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
On a hybrid laptop (issue #3) Indium took whichever Vulkan device the
loader listed first, usually the integrated GPU, and still created a
VkDevice on every other supported GPU, waking a discrete GPU it never
used. Drawables reach the presenting OpenGL context as opaque FDs, which
only work on one physical device, and nothing checked that.
Indium now appends the device list, the selected device and the reason
to guest /tmp/metal-device.log, and creates a VkDevice for the selected
device only (/tmp/metal-all-devices restores the old behaviour).
/tmp/metal-gpu or METAL_INDIUM_GPU selects a device by index or name
for diagnosis. QuartzCore logs the presenting GL renderer and compares
its device UUID with the Vulkan device's (match=yes / MISMATCH); it
queries Vulkan itself, so it needs no new libindium export.
On the RX 6600 desktop the game selected RADV, created 1 VkDevice of 2
supported (lavapipe no longer gets one) and logged match=yes in a match.
Running on NVIDIA through prime-run is documented but not validated.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Kalah