fix: build and test on Windows
- .gitattributes keeps the session's scripts LF under core.autocrlf=true;
CRLF scripts made every session exit with status 2.
- runtime_inputs.py and the scrub record / log zip use '/' in relative
paths, which Windows reported with '\' (runtime inputs never matched).
- The read-only-folder scrub test is skipped on Windows, which still
creates files in a read-only directory.
Co-authored-by: xXJsonDeruloXx <danielhimebauch@gmail.com>
* fix: redist markers a replaced registry lost are seeded again
droiddeck-seed-redists recorded what it seeded in .droiddeck-redists and
trusted that record from then on. A Proton update replaces its
default_pfx system.reg but leaves our record beside it, and every new
prefix copies both, so new games got a record of markers their registry
never had: Steam then ran their install scripts for real. Batman: Arkham
City's AMD Dual-Core Optimizer installer hangs there, leaving the game
on "Running install script" on every launch.
Each recorded marker is now looked up in the registry, in both views
where both were written, and seeded again when it is gone; a record of
markers nobody wants any more is rewritten without them.
* Check every VC runtime architecture before keeping the vcrt marker
vcrt writes Installed for x86, x64 and arm64 in both registry views, but
present() only looked at x86. If only an x64 or arm64 value was gone the
runtime would not be seeded again. Require all six values and test the
loss of the arm64 Wow6432Node one.
Arkham City (BatmanAC.exe) stopped with Unreal's "Fatal error!" about
1.5 s after start: an access violation in BatmanAC.exe right after
steam.dll loads. The game leans on x87 and needs full precision under
FEX (FEX wiki), while the default FEX preset sets
FEX_X87REDUCEDPRECISION=1. Engine fixes sit above the shared preset and
below the player's own game profile, so a fix keyed on BatmanAC.exe
beside the started exe (BatmanAC.exe or BmLauncher.exe) restores
x87 precision for this game only. Arkham Asylum, which shares
BmLauncher.exe but not BatmanAC.exe, keeps the preset.
install() replaces an existing cache entry by calling detach(old), which
destroys the listing when no reader holds it, and then reads old->wd to
decide whether to release the watch. When the old listing has no readers,
that read happens after the free.
Two threads that enumerate the same uncached directory at the same time
reach this path: both build a listing, the first installs an entry with
no readers, and the second replaces it.
Save the watch descriptor before detaching, as drop() already does.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Both store modules are rewritten from a protocol-level spec: new Rust engines behind a small JNI surface (nativeCall/nativeInstall) and thin Kotlin facades, with the launcher's own user agent.
Epic: sign-in carries over, library, manifest download with byte-exact verification, assembly, uninstall, launch arguments, achievements.
Amazon: sign-in carries over, library, manifest download with per-file hash checks, path checks, executable detection from fuel.json, uninstall.
On-disk install layout is unchanged, so existing installs keep working.
Store game pages now take controller focus on their lead hero action (Install, Uninstall, Pause, Resume, Get, Buy), fall back to Back, and keep D-pad focus inside the page while it covers the library.
Space checks before Epic, GOG and Amazon installs (measured zero vs unknown), clear failure reasons with retry/clear, resume-aware checks, and safe per-file replacement so a failed or cancelled update never breaks an installed game.
Epic installs stream chunks straight into staged files in parallel (most bytes written once, a small budgeted cache for shared chunks, about 1x game size instead of 2x), pace updates by free space, and keep the lookahead, batch and bounded Java writers as hidden alternatives.
Co-authored-by: Juan Marti Mercado <juanma@microsoft.com>
Co-authored-by: The412Banner <the412banner@users.noreply.github.com>
* fix: desktop session: Shift no longer stays held after typing capitals on a physical keyboard
Android repeats a held key as more key presses. Those were forwarded to the
session as more wl_keyboard presses, and KWin's xkb counts every press of a
modifier, so Shift held for a few capitals took several presses and one
release and stayed held: digits typed !@#, letters stayed capital. The same
applied to Ctrl and Alt.
- the compositor no longer sends a press for a key that is already down;
releases always go through (clients repeat keys themselves from the
repeat_info it already sends)
- the app stops forwarding Android's repeat events for keyboard keys
Reported by whatsupbros with a Keychron Q7 over USB - thank you.
* Remove stale comments from statx patch
---------
Co-authored-by: The412Banner <the412banner@users.noreply.github.com>
Co-authored-by: xXJsonDeruloXx <danielhimebauch@gmail.com>
droiddeck-msi-install: a DuplicateFile entry onto the file itself (dotnet.exe in the .NET host MSI) is skipped like Windows Installer does, instead of stopping the whole install with SameFileError. Fixes .NET Desktop 3/6/7/8, .NET Core 3 and .NET 5. Thanks to whatsupbros for finding and tracing it.
* Add opt-in zRAM compression for the Steam client while a game runs
A "zRAM compression" toggle under Steam > Settings > Session, below
"Keep downloads running". Off by default, kept once set, and disabled
on devices without swap or with a kernel older than 5.4.
When on, the session watches for Steam's reaper. 20 s into a game the
client and its web helpers page their private anonymous memory out to
the device's zram with MADV_PAGEOUT from inside their own process
(libblsession zram.c, SIGURG), again every 5 minutes, and read every
swapped page back once the game exits. No root is needed: only a
process itself may do this without CAP_SYS_NICE. With the toggle off
nothing is installed or started.
RedMagic NP06J, Palworld: Steam's resident memory fell from ~1.7 GB to
~1.1 GB with ~560 MB in zram, available memory during play rose by
~0.5-0.8 GB, and ~2 MB was left in zram after the read-back.
* Turn zRAM compression on by default below 12 GB of RAM
Until the toggle is changed, it follows the device: on when the kernel
supports it and the device has less than 12 GB of RAM, read from the
advertised memory size on Android 14+ and from total memory before that.
Turning it off is kept.
---------
Co-authored-by: xXJsonDeruloXx <danielhimebauch@gmail.com>
* feat: add Cardboard mode with side-by-side lens-distorted eyes and 3DoF head tracking
* fix: derive the Cardboard field of view from the panel and recentre fully when far from upright
* feat: share the Cardboard head pose with the guest and add a SteamVR driver for it
* feat: run SteamVR's compositor on Turnip/KGSL with a display-extension layer and a driver-direct HMD
* feat: show SteamVR frames in Cardboard mode through a driver-side presenter and a fake direct-mode display
* feat: install and start SteamVR for Cardboard mode from the session drawer
* docs: describe Cardboard mode and the SteamVR path
* feat: drive SteamVR hands from the game pad and keep the Cardboard headset out of standby
* docs: record where the SteamVR dashboard stops
* fix: map Cardboard head tracking to the screen's axes and keep turns about the real vertical as yaw
* fix: give Cardboard its previous picture back when SteamVR stops
* fix: report sem_otime from emulated SysV semaphores so SteamVR's web helper can start
* fix: bridge SteamVR's shared semaphores across processes and keep Steam alive while SteamVR runs
* feat: present SteamVR straight to DroidDeck's compositor at panel resolution
* docs: update Cardboard VR notes for the dashboard, semaphore bridge and direct presenter
* fix: keep Wine and FEX signal handlers in the VR layer; add semaphore tracing
* fix: clear a stale SteamVR active marker on status
* fix: clear only an old SteamVR active marker
* fix: unblock SteamVR clients on Turnip and keep Steam's web helper alive across SteamVR runs
* fix: judge shared waits before a submit's own signals and on Vulkan 1.0 devices
* docs: record VR games, SteamVR Home and OpenXR results on the Thor
* docs: add Cardboard VR learnings from the Thor bring-up
* docs: add Cardboard VR screenshots from the Thor
* fix: keep the VR layer's display stand-in to SteamVR runs and clear a leftover marker at session start
* fix: report SteamVR start and stop failures from the helper's exit code
A game that closes as soon as it starts can be tripped by Proton's Steam
setup (steam.exe, the Steam registry entries, the Steam client bridge).
The Games page of the session drawer gets a per-game Start without Steam
switch, stored as DROIDDECK_NO_STEAM=1 in the game's environment and listed
in its profile summary.
droiddeck-game-env acts on it for a game launch only: a Proton with umu's
launch (GE-Proton, Proton-CachyOS, DroidDeck-Proton) gets UMU_ID, so it
runs umu.exe in place of steam.exe and still sets up the prefix; a Steam
game keeps its protonfixes under its own id, an added game gets 0. Any
other Proton runs the game straight in the prefix (runinprefix).
* feat(profiles): add persistent per-game compatibility profiles
* fix(profiles): show Windows component overrides
Include per-game Windows component selections in the resolved profile summary so the session overlay marks those games as custom and shows the selected components.
* fix(session): detect Proton games across runtime paths
* fix(profiles): select the active Gamescope game
* fix(profiles): improve per-game reset feedback
* fix(profiles): align session components with game overrides
* fix(session): prefer selected Steam game during refresh
* fix(session): resume profile following when reopening games
* chore: keep local dev testing outside the repository
* test(session): remove local dev package path
* fix(session): preserve profiles when refresh fails
* fix(session): reject empty game profile reset scope
* perf(session): cache drawer inventory by refresh revision
* fix(profiles): align game presets and profile reset behavior
* fix(profiles): preserve originals and normalize shortcut ids
* fix(profiles): keep editor scope stable after rebase
* fix(tests): host Compose tests in both build variants
* fix(profiles): recover interrupted component selection saves
---------
Co-authored-by: Kurt Himebauch <136133082+xXJSONDeruloXx@users.noreply.github.com>
* fix: Red Dead Redemption 2 launches again, its Vulkan runtime installer is left to run
droiddeck-seed-redists marked every process in every game's install script
as already run, Red Dead Redemption 2's VulkanRT-1.1.108.0 installer
included. Steam's compat config for the game says nativevulkanloader, so
Proton loads vulkan-1.dll native only, and Khronos' loader is in the prefix
only once that installer has run. Without it the Rockstar launcher starts
RDR2.exe and it exits with STATUS_DLL_NOT_FOUND (0xc0000135) before it
opens a window.
A Vulkan runtime installer is now left out of the seeding, and a marker an
earlier seeding wrote for one comes out of each prefix and Proton template
once (dates kept), so the client runs the installer on the next launch.
The installer finishes in a few seconds under FEX.
* fix: Steam's screensaver no longer covers a running game
The client's defaults show screensavers while a game runs, after 300 s on
AC power without a button press - and a held stick or trigger is no press.
Red Dead Redemption 2's intro runs longer than that: the slideshow covered
the game, and the button that woke it reached the game too (B asked to
quit to the desktop). Customization/Screensaver/WhileInGame is seeded "0"
in config.vdf once, like the UI scale and native resolution, so turning it
back on in Settings stays the user's choice.
On the RedMagic: 18 minutes from the last button press through the intro
and riding, focus never left the game and no screensaver was created.
* fix: Rockstar's hidden launcher no longer takes cores from its game
The Rockstar Games Launcher keeps drawing its Chromium interface
(SocialClubHelper.exe: renderers, and a GPU process compositing through
DXVK) for as long as the game runs, hidden behind it - and Wine passes
Chromium's priorities on, so those threads ran above the game's (the GPU
process at nice -10), across every core. With Red Dead Redemption 2 in
front that was 1.3 to 2.5 cores under FEX for a page nobody sees.
droiddeck-launcher-cores follows gamescope's focus from the agent's
focus.json: while a game window has it, every SocialClubHelper thread runs
on the last core of the slowest cluster at nice 19; when the launcher's own
window has focus again each thread gets its own mask and priority back
(also on exit). Stopping the helpers would not hold - proot resumes a
stopped tracee. Started for Steam sessions; BL_LAUNCHER_CORES=0 turns it off.
RedMagic, RDR2 all Low, riding at the thermal limit, same spot and session:
released 46.5 fps, 1% low 27.5, p99 31.8 ms, CPU 89%, 2200 MHz;
held 48.7 fps, 1% low 32.6, p99 27.6 ms, CPU 79%, 2638 MHz.
* fix: sessions clean up what earlier ones left in /tmp
The guest's /tmp is a directory in the rootfs that no boot ever empties,
and Android ends sessions by killing them. Every session left a
/tmp/mangohud.XXXXXXXX (mktemp, never removed), and the Steam client, its
web helper, Chromium apps, dbus-run-session, AppImage runs and KDE their
own temporary files and sockets: 1,422 entries on the RedMagic.
mangoapp's config is now one fixed file rewritten each session, and the
session's first pass sweeps each kind of leftover while nothing that could
own one is running, with a single find. On the device that cleared 1,391
entries over two sessions, leaving 38.
Setup > Proton versions lists DroidDeck-Proton (The412Banner/DroidDeck-Proton) beside
GE-Proton and proton-cachyos: same aarch64 tarball pattern as GE, verified against its
sha512 file, installed by droiddeck-proton-extra (also as "droiddeck" in the session queue)
and adopted by steam-compatibility like any other extra, so it shows as
"DroidDeck-Proton <ver> (Compatible)". Its folder has a space; nothing in the install,
adopt, esync or components paths splits on it.
Update checks: each source's newest full release is looked up at app start and when the
page opens, at most every 6 h (offline keeps the last answer). A newer release shows
"Update available -> <tag>" with Update; it installs beside the old build, which stays
until removed ("Remove old ... versions"), since games are mapped to it by name.
Installed builds are ordered by their version numbers (11-9 before 11-10), not by name,
and the tag each was installed from is recorded. Prereleases and drafts are skipped.
Each game's Windows components now come from the list its own source
trusts, and the ones that list names turn on by themselves. The user's own
on/off switches always win; each automatic row says why it is on.
- Steam games: the Steamworks Shared redistributables Steam installs with
the game (appmanifest SharedDepots, else the client's appinfo.vdf).
Steam keeps its install scripts marked as already run.
- GOG: the build manifest's dependencies. Epic: the manifest's
prerequisite (Unreal's maps to VC++ and DirectX). Amazon: fuel.json
PostInstall. Older GOG/Epic installs get their list fetched once with the
saved sign-in; game files are never downloaded again.
- Custom games (the + button, Add all, migrated folders): one shared Steam
match per game, also used for artwork. An appid from the game's files
turns components on; a name-only match stays Recommended; no match falls
back to the folder scan.
- Steam app info for apps the client never saw comes from an anonymous
Steam PICS lookup (CM websocket, anonymous logon, product info), cached
per appid; the user's Steam sign-in is never used.
- Missing components download in the background on one low-priority
worker that pauses during sessions and waits on battery saver, heat or a
metered network unless the user asked. A game Steam installs during a
session is readied right away so it can be played without ending the
session; the launch waits up to 10 s only for a component still on its
way.
- Page: "On for this game" with reasons, a quiet "Checking what this game
needs..." only for games never worked out, rows that ease in, pieces that
need a download greyed out while offline. The intro names no other app.
- The folder scan skips Unity's *_Data/Managed (Mono, not .NET Framework).
Preload the FEX fault report only where x86 loaders look
droiddeck-fex put the x86 libfaultreport.so in LD_PRELOAD, which FEX's own
arm64 loader reads too - and FEXServer's, and every child FEX's - so each
x86 start printed "ERROR: ld.so: object '.../libfaultreport.so' from
LD_PRELOAD cannot be preloaded", and an i386 child added "wrong ELF class".
It is now listed in the x86 runtime's own /etc/ld.so.preload, which only
the loaders inside FEX read, as /usr/$LIB/droiddeck/libfaultreport.so.
faultreport.c builds for i386 too, so 32-bit programs get the report as
well instead of an error.
* Run programs, scripts and AppImages from the desktop
Windows programs open with the Proton Steam has chosen, and a Steam game's own
launch program is handed to Steam. x86-64 Linux programs and game scripts run
under FEX, and an exec the kernel refuses goes to droiddeck-open through
libblsession's binfmt.c. Programs on noexec shared storage pass KIO's checks; a
folder of Linux programs there, and sharun AppImages, are added by the app once
and run from the copy. Executable scripts run without KIO's prompt, mpv plays
on the software desktop, FEXServer's socket moves under XDG_RUNTIME_DIR, and
Proton runtimes no longer show up as games.
* Switch to Desktop from Big Picture without Deck mode
Big Picture's Switch to Desktop asks the client's SteamOS manager, which the
client only has when started as SteamOS (Deck mode). Without it the request
failed and the screen stayed on "Switching to Desktop..." for good. The live
compat helper now sees the request on its way to the client and hands it to
steamos-session-select, which opens the KDE desktop as Deck mode already did.
* Run x86 programs from shared storage where they lie
FEX asks for the guest's PROT_EXEC when it maps a program or library, which
the kernel refuses on noexec shared storage; FEX never runs those pages
itself, so libblsession maps them without it in FEX's processes. An x86
game, its libraries and its start script now run in place instead of being
copied in; only native code on noexec storage is still copied.
A Steam library's games are never copied, and a Linux game's launch program
opened from the desktop goes to Steam like a Windows one.
* Front end: draw the capsule wall without clipping it every frame
The tilted wall clipped each capsule to a rounded rect and recomposed its columns on every drift
step, and a rotated rounded clip is rasterised in software on the render threads: the idle main
page held about 1.7 cores. The art is now loaded at the capsule's size with its corners already
rounded, each column draws its capsules itself in one layer moved by the drift, and the drift
advances at most every 15 ms. The idle page takes about a third of a core.
* libfakeinput: let calls on descriptors it does not serve skip its lock
Every read, write, ioctl, close and poll of every process in a Proton session went through the
controller lock and the three map lookups, Wine's wineserver pipes and the game's files alike.
A bitmap of the descriptors the maps hold, set before one is handed out and cleared when it is
closed, lets everything else go straight to the kernel.
* Pad motion: ask the display for its rotation four times a second
The rotation was looked up for every sensor reading, every 2 ms, and cost more than the rest of
the reading; it is now refreshed every 250 ms, and the readings no longer allocate pairs.
* Session: keep the library passes out of a game's launch
The redist seeder stamped each prefix with the checksum of the whole set of names, which differs
between the session's pass and the game launch's (the game's own names are added), so every
launch re-appended the markers to every prefix in the library while the game loaded. Each
prefix's marker now lists the names it was given and only the missing ones are added.
The beat's seed and compatibility passes, and the Lossless Scaling ownership probe (which loads
steamclient.so), also wait while a wineserver runs: a launch rewrites the manifest and made them
due at exactly that moment. The game's own launch seeds its prefix regardless.
* Agent: keep the focus watcher's X connection open while nothing changes
The connect timeout stayed on the socket, so with no property change the read timed out, and the
watcher dropped the connection, reconnected and took a fresh snapshot every few seconds.
* Game environment: start Cyberpunk 2077 without REDlauncher
The prelauncher opens REDlauncher, a Qt WebEngine window that hands over to the game half a
minute later and then stays up beside it, well over a core busy for as long as the game runs; its
own --launcher-skip started the game twice. The game's executable is started directly instead.
* netif: answer Wine's network driver from a reading of the link
Mortal Kombat X has Wine's network driver enumerate every interface, address and route table
hundreds of times a second. Each enumeration read /etc/droiddeck-net several times, asked the
kernel for its interface dump again (a netlink socket bound under proot, refused each time) and
looked the real functions up with dlsym. The link file is now read at most once a second, a
refused dump is not asked for again within that second, and the real functions are looked up
once.
* Fast path: create memfds without a stop in proot
proot looks at a memfd's name only to refuse the few that are then executed or locked; every other
memfd_create is now issued from the trampoline. The session's /proc/net and status copies made
one per read, a stop each. Without memfd in the kernel the call still goes to proot's ashmem.
* Logs: run each redaction rule only on a line that could match it
A session's end copies some 70 MB of the Steam client's logs and puts every line through all
twenty-odd rules, most of a minute of a core on the phone after every session. Each rule now runs
only on a line holding what it cannot match without: its keyword, the five separators of a MAC,
the four dashes of a GUID, a long enough hex or token run, the learned account or address text.
Lines with non-ASCII text still meet every case-insensitive rule. Over 895,000 lines of real
client logs, with the device's accounts and addresses learned, the output is identical and the
pass takes a twentieth of the time.
Fix JNI local reference leak in createSubprocess
The argv and envp loops took one local reference per element from
GetObjectArrayElement and never released it, so the references piled up
until the native call returned. ART aborts the process at 512 local
references ("local reference table overflow"), and both loops share the
table, so a long argument list plus a large environment could hit it.
Delete each element's local reference once its chars are copied and
released, after ReleaseStringUTFChars since the chars depend on the string.
Also release the cwd chars against cwd rather than cmd; the call passed
the wrong jstring.
Adds a host-side test that runs termux.c against a mock JNIEnv that counts
local references and checks release pairing. It needs a jni.h (JNI_INCLUDE,
JAVA_HOME or the NDK) and skips without one.
Fixes#507
The publish workflow's push trigger fired again on the branch's rebase and overwrote the
release asset with a freshly compressed copy, so main's first build failed the checksum and
fell back to the registry pull. The env file now pins the asset as published; the workflow is
dispatch-only, defaults to a new tag, and refuses to replace an existing asset unless asked.
* CI: host the Debian image the preloads build in on a release of this repo
The preloads step pulled debian:bullseye from a public mirror on every build whose preload
cache missed; the mirrors rate-limit the shared runner addresses and failed two builds in an
hour tonight. Debian 11 is frozen upstream, so the image is published once (workflow
ci-debian-image.yml, release ci-debian-bullseye-r1, pinned in tools/linuxfs/debian-image.env)
and the build downloads that file; only if that fails does it pull, with retries.
* CI: pin the published Debian image (ci-debian-bullseye-r1, 34 MB)
* CI: a new Debian image rebuilds the preloads (env file joins the cache key)
XxJSONDeruloxX