Cemu's own Wolf-facing controller slots (up to 4) still each need a
separate manual bind through Cemu's Input Settings even after
./manage.sh controllers makes them distinct SDL devices — that dialog's
Save button is documented elsewhere in this README as getting cut off
the viewport, making binding all 4 slots by hand painful.
Validated the underlying assumption against two real, separately
hand-mapped profiles pulled live from a working install (Nintendo
Switch Pro on controller0.xml, Xbox One S on controller1.xml): both
contain the exact same 24 <mapping>/<button> pairs, just in a different
order — real proof Cemu's mapping format is device-agnostic and
order-independent, not just the README's prior unverified claim.
New command clones a proven-working <mappings> block onto a new
device's controllerN.xml, needing only that device's own real <uuid>
(which still has to come from Cemu itself — a hand-computed SDL GUID
risks not matching what SDL actually reports for the live device, so
this never guesses one). Getting that uuid only needs a single-button
minimal bind in Cemu's UI, not full mapping, since only the uuid gets
kept from it.
Verified end-to-end against the real captured controller0.xml/
controller1.xml content: cloning controller0's mappings onto
controller1's real uuid reproduces the exact same 24-pair set Cemu
itself wrote, and the output validates as well-formed XML.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Copyable follow-up for the existing Cubeb/PulseAudio hang note: checks
whether a full Wolf restart has actually happened, whether PulseAudio
inside the Desktop container has any sinks at all, and pulls Cemu's own
log.txt plus its current <Audio> settings.xml block — narrows "stuck on
Disabled" down to a container-level audio problem vs. Cemu's own
device-switch path before guessing at a fix.
Verified the heredoc escaping by rendering the write_readme block through
a real bash heredoc and confirming $WOLF_DIR interpolates while every
other $ stays literal in the output.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Root-caused against the user's own live output: the settings file had
<string name="ROMDirectory" value="" /> — the key was present, just
empty. The self-healing check only tested whether the string
"ROMDirectory" appeared anywhere in the file, so a key that exists with
a blank value (ES-DE can write this itself if its own first-run "select
ROM directory" step goes unanswered in a headless Moonlight session,
saving an empty path back over GOW's template) looked "already
populated" and got skipped — leaving ROM discovery broken even after
the fix had run.
Now ROMDirectory is independently forced to /ROMs whenever its value is
blank, on top of (not instead of) the existing RunInBackground patch —
no longer gated on the key's mere presence. Verified against the exact
reported bug (ROMDirectory present but blank) plus the existing missing-
file, already-correct, and idempotent-rerun scenarios in isolated /tmp
harnesses before touching the real script.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Adding an emulator (Cemu, etc.) to Steam as a non-Steam game previously
needed Steam's own Big Picture file-browser flow through Moonlight. This
writes the shortcut directly into Steam's binary shortcuts.vdf instead,
matching this repo's existing no-manual-wizard pattern (ge-proton,
install-ea-app). Motivated by wanting to test whether Steam Input's
per-device controller tracking (distinct device paths, not SDL GUIDs)
can hand Cemu 4 explicitly-assigned controllers where ES-DE/Cemu's own
SDL-based handling can't tell identical controllers apart.
- New generic binary VDF (KeyValues) parser/serializer: round-trips any
existing shortcuts.vdf entries byte-for-byte and only inserts/replaces
the one entry matching the given Exe path, so it's safe against a file
that already has real, hand-configured shortcuts. Verified in isolated
/tmp harnesses: fresh file, idempotent re-add, a second distinct entry,
and preserving a synthetic pre-existing GUI-set entry (icon,
LaunchOptions, tags) untouched.
- Steam's own CATALOG entry had no mount for emulators/ at all (unlike
esde/retroarch) — added emulators:/home/retro/Applications so an
AppImage is actually reachable from inside the Steam container. Like
the ES-DE settings mount, this only takes effect on a freshly created
WolfSteam container (Wolf reuses existing ones) — noted in the README.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Previous version only patched an already-populated es_settings.xml,
requiring a manual "connect via Moonlight once" step before the fix
could apply — the user correctly pointed out the script can't know
whether that step has been done, so a manual dependency isn't a real
fix. This embeds GOW's own real es_settings.xml template (fetched from
games-on-whales/gow, RunInBackground pre-flipped false) and writes it
directly whenever the on-disk file is missing or is the old broken
stub (detected by the absence of ROMDirectory, which only a real
GOW/ES-DE-written file has). An already-populated file still only gets
the one RunInBackground line patched, leaving everything else alone.
Verified against all 4 scenarios in isolated /tmp harnesses before
touching the file: missing file, the exact broken stub from the user's
regression, GOW's real full template, and idempotent re-runs.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live on a real install: after the previous commit, ES-DE found
zero games on every system. Root-caused by walking GOW's own
apps/es-de/build/scripts/startup.sh line by line: GOW's image bakes a
FULL es_settings.xml template (100+ settings, including
ROMDirectory=/ROMs, confirmed against the real template file in the
games-on-whales/gow repo) and its own startup script copies that
template into place exactly once, the very first time the container
starts, gated on the settings file not already existing.
The previous fix pre-created a minimal stub file (containing only
RunInBackground) before the container had ever started - GOW's startup
script saw that file "already there", skipped writing its own template
entirely, and ES-DE fell back to vanilla upstream defaults for
everything else in that file, including ROMDirectory, which is not
/ROMs by default. Every system's games vanished as a result.
Also corrects an earlier wrong claim in both the code comment and
README: ES-DE's own *compiled* default for RunInBackground is false, but
GOW's own template deliberately sets it to true - that's the real,
confirmed source of the original background-launch bug, not an
accidental toggle as previously assumed.
Fix: only ever patch an es_settings.xml that already exists with real
content (`[ -s "$_ESDE_SETTINGS" ]`) - never create one. On a genuinely
fresh install, before EmulationStation has ever launched once, the
installer now says so explicitly and defers rather than guessing, with
clear instructions to connect once and re-run. Verified against GOW's
actual real template: the patch flips just RunInBackground (line count
unchanged, ROMDirectory and everything else untouched), is idempotent on
re-run, and the skip-when-absent gate was verified directly against both
the missing-file and populated-file cases.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Live user feedback: being asked to pick between 6 numbered entries that
were all the exact same device (client_id repeated 6x from re-pairing)
was genuinely confusing, especially right when the user was already
trying to get to the controller-type poll further down the flow.
The picker now dedupes to distinct client_ids only, first-occurrence
order, tagging a collapsed entry "(paired Nx)" - matches what Wolf's own
get_client_by_id() actually resolves to anyway (first match for a given
id), so nothing is lost by not offering the later duplicates as separate
choices. Verified against the user's real 8-entry (6 duplicate + 2
unique) client list, both with and without an active session, confirming
the controller log-poll still runs correctly right after selection in
both cases.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Fixes a real, confirmed-live, cross-system bug: with ES-DE's "Run in
background (while game is launched)" enabled, ES-DE keeps listening to
every controller even after a game/emulator takes focus - a second
controller's input can reach ES-DE's own menu and launch a completely
different game underneath whatever's already running. ES-DE's own
USERGUIDE.md names this exact failure mode.
ES-DE's compiled default for this setting is already off (confirmed
against its real Settings.cpp), but nothing in this repo asserted that
durably, and ~/ES-DE (settings, gamelists, scraped media, logs) had no
mount onto the game drive at all - confirmed against ES-DE's own source
(getAppDataDirectory() is a plain $HOME/ES-DE) and GOW's own es-de
startup script. Wolf normally reuses the same app container across
sessions rather than recreating it each time (confirmed against Wolf's
docker.cpp - it only removes the container on exit if
WOLF_STOP_CONTAINER_ON_EXIT=TRUE, which this repo never sets), so this
mostly didn't bite day to day, but it meant the setting - and everything
else under ~/ES-DE - wasn't safe across an actual reinstall the way
roms/saves/BIOS already are.
Mounts esde-settings/ onto ~/ES-DE/settings specifically (additive,
doesn't touch the existing custom_systems mount) and writes a surgical,
idempotent RunInBackground=false into es_settings.xml on every install -
never a full rewrite, since that file is the user's own complete
settings state. Verified the write logic against fresh-file, existing-
file-with-other-settings, and flip-an-existing-true-to-false cases, plus
idempotent re-runs.
Gamelists/scraped-media durability under ~/ES-DE is a related, noted gap
- not fixed here.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Two follow-ups from live testing that just confirmed the whole
controllers_override chain works end to end:
1. manage.sh controllers now polls `docker compose logs wolf` for
"Creating <TYPE> joypad for controller <N>" lines (most recent per
slot wins) and shows what it last saw before asking anything, using
it as the suggested default at each prompt. Every prompt is now
labeled "1st controller"/"2nd controller"/etc - always meaning
Wolf's own 0-indexed "controller 0"/"controller 1" - instead of
asking for a raw slot number, closing the exact client-vs-controller
0-vs-1 confusion surfaced live earlier. Documents plainly that this
can show what TYPE a slot last used but can't show or control WHICH
physical controller becomes which slot - that's decided by Moonlight
client-side, upstream of Wolf's API entirely.
2. Documents a full recovery procedure in the Wii U section for when
Cemu's Input Settings Save button is cut off-screen (confirmed live:
happens even maximized, at higher resolutions, and closing the
window doesn't save) - an Alt-drag/Alt+F7 window-manager fix first,
then editing Cemu's own controllerProfiles/controllerN.xml directly
on the host (plain XML, not inside the container) as a full GUI
bypass. Documents the key insight that made this actually work live:
<mapping>/<button> pairs are universal SDL_CONTROLLER_BUTTON_*
semantics (confirmed earlier against real captured profiles from
different controller brands), so a working mapping can be copied
verbatim into a different device's empty <mappings/> block without
redoing the live button-capture in Cemu's GUI at all.
Verified the redesigned manage.sh controllers flow end to end against a
mocked docker compose logs matching the user's real log output (both
the log-seen and no-logs-seen paths), plus the ordinal-suffix logic in
isolation.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Live user feedback: "Exactly one client is actively streaming right
now" sitting directly above "How many controller slots" read as "only
one controller is connected" - an easy misread since "client" (a paired
Moonlight session/device) and "controller" (a physical gamepad slot
within that session) are conflated by ordinary English but mean
completely different things here. The tool has no way to detect
controller count at all; it only ever picks which Moonlight
session/device to configure.
Reworded every message in this flow to say "Moonlight session/device"
instead of bare "client", and added an explicit line before the
controller-count prompt clarifying it's a separate question the user
has to answer themselves. Verified the full flow still runs correctly
end to end against the same mocked duplicate-heavy client data used
earlier.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live: the paired-clients list is a wall of meaningless
cert-derived numbers with nothing human-readable attached (verified
against Wolf's own PairedClient API schema - no name, no IP field
exists there at all), so there was no way to tell which entry was the
user's actual device. Made worse by Wolf not deduping repeated pairings
of the same device (the user's real box showed the same client_id 6
times).
Fix: cross-reference against GET /api/v1/sessions, which DOES carry
client_ip for every currently-streaming session, and tag each paired
entry as "ACTIVE - streaming from <ip>" when it matches one. When
exactly one client is actively streaming, it's now auto-selected with
no prompt - the practical case for a single-user setup. Falls back to
the manual picker when nothing's active or multiple clients are (now
showing which IP is which, so the choice is no longer a guess). Verified
against the user's real duplicate-heavy client list plus mocked active-
session data covering: single active session (auto-select), no active
sessions (manual fallback), and two distinct active sessions (manual
fallback with both IPs shown).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live: a plain-user curl against /var/run/wolf/wolf.sock failed
with "Immediate connect fail ... Permission denied". The socket is
srwxr-xr-x root:root - Docker auto-creates the host-side /var/run/wolf
bind-mount source as root, and Wolf itself runs as root in its
container, so group/other never had write permission on the socket file
(required to connect() to a unix socket, not just read/traverse it -
`ls -la` on the containing dir working fine was a red herring).
Fix: sudo on both curl calls, matching how every other root-owned Wolf
state file ($WOLF_CFG) is already touched elsewhere in this script via
`sudo python3 -`. Documents the sudo prompt in the README too.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Spells out what the three controllers_override values actually create
(Xbox One, PS5 DualSense, Switch Pro Controller - confirmed against
inputtino's own source, not just the enum names) instead of leaving the
reader to guess from bare AUTO/XBOX/PS/NINTENDO tokens, plus a note that
forcing a non-matching type only affects on-screen button-prompt
artwork, not functionality.
Adds a 4-controller section: Cemu's own Input Settings goes up to 8
slots and Wolf's wire protocol has no hardcoded 4-controller cap
(confirmed against control/input_handler.cpp), so the ceiling in
practice is the Moonlight client's own limit. With only 3 concrete
forced types available, a 4th controller can't get a guaranteed-unique
GUID - documents the one-at-a-time Cemu binding workaround for that
case, flagged as reasoned but not confirmed live.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Root cause (confirmed against inputtino's own uinput device-creation
source, and against Wolf's real virtual-pad architecture): every
virtual gamepad Wolf creates of the same type gets an identical SDL
GUID, since a GUID identifies a controller model, not a physical
instance - the same behavior two real identical controllers would have.
This isn't a Cemu bug specifically; it's why any app whose own
controller picker keys off GUID (Cemu confirmed live) can't reliably
tell two same-type Wolf pads apart, driving every player from
controller 1.
Wolf's own per-client `controllers_override` setting sidesteps this at
the root by forcing each controller slot to a genuinely different pad
type (e.g. slot 1 = Xbox, slot 2 = PlayStation), so their vendor/product
IDs - and SDL GUIDs - actually differ. `./manage.sh controllers` drives
this via Wolf's real REST API (confirmed against its OpenAPI schema,
not its docs page's own stale example command) rather than hand-editing
config.toml: lists paired clients, prompts for a pad type per slot, and
POSTs the update. Needs Wolf's API socket exposed to the host
(WOLF_SOCKET_PATH + a /var/run/wolf bind mount, matching Wolf's own
documented pattern), added to the wolf service's docker-compose.yml.
Full flow verified with a standalone test harness (mocked curl against
real API response shapes, a real AF_UNIX socket file) covering: single-
vs multi-client selection, invalid selection, invalid slot count,
lowercase/unrecognized type input, and the missing-socket path.
Also documents live user feedback that AntiMicroX (added last session)
is confirmed not working for TI-99/4A - flagged in the README rather
than left silently unverified, with this fix noted as the actively
maintained path for the multi-controller problem instead.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
The PCEE2 libretro core (a separate, actively-developed PCSX2 port,
WizzardSK/pcee2-libretro) is only in ES-DE's own bundled es_systems.xml
as of 3.5.0, which hasn't been released yet - confirmed against ES-DE's
own CHANGELOG.md and by diffing the v3.4.1 tag's real ps2 system block
against git master's (identical except for this one missing <command>
line). ghcr.io/games-on-whales/es-de:edge always installs the latest
*released* AppImage (confirmed against gow's own Dockerfile, which
queries GitLab's releases API), so the container this repo runs
genuinely has no PCEE2 entry at all yet - not a missing core download,
a missing menu entry.
Backfills it via the same esde-custom-systems/es_systems.xml mechanism
already used for the TI-99/4A and Wii U customizations, replicating the
real v3.4.1 ps2 block verbatim plus the one new command (as the default,
matching upstream's own placement). The core file itself needs no extra
handling - confirmed live against the libretro buildbot that
pcee2_libretro.so.zip is already covered by the normal RetroArch cores
pre-download.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
AntiMicroX maps gamepad buttons to synthetic keyboard/mouse events -
useful here because ti99sim-sdl's own joystick handling only ever emits
digit keys 1-9 for a raw button (confirmed against its source), with no
path to 0/Enter/Q/Esc, and because it's a plausible angle on Cemu not
reliably telling apart two Wolf virtual pads that share an SDL GUID.
Scoped to just these two ES-DE systems via a second, separately-labeled
"(AntiMicroX)" <command> alongside each one's existing default - ES-DE's
own multi-command "alternative emulators" mechanism - rather than
touching every system. Each launches through a small wrapper script
that starts AntiMicroX hidden (--no-tray --hidden --eventgen uinput,
flags confirmed against AntiMicroX's own commandlineutility.cpp source)
against a profile the user builds themselves via AntiMicroX's own GUI in
the Desktop/XFCE app, execs the real emulator, and kills AntiMicroX on
exit.
AntiMicroX needs /dev/uinput to inject events under ES-DE's Sway/Wayland
session (its XTest backend needs Xwayland, not present here), so the
esde catalog entry now requests it via both GOW_REQUIRED_DEVICES and a
real device grant - and make_app_block/the reinstall-update path both
learn to read/refresh a per-app 'devices' field instead of always
emitting an empty array, so this actually reaches an existing install on
a rerun, not just a fresh one (verified live against a synthetic
pre-uinput-era config.toml).
Also fixes a latent whitespace-eating bug in every custom_systems
strip-and-reappend regex in this file (trailing \s*\n? reached into the
next sibling block's leading indentation once two custom systems
coexist in the same file) - cosmetic only, but it would have made every
rerun churn indentation between the ti994a and wiiu blocks. Caught and
verified via a standalone test harness exercising all four
strip-and-reappend blocks together across repeated passes.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Joystick buttons beyond the first only reach digit keys 1-9 in
ti99sim-sdl's own source (src/sdl/ti994a-sdl.cpp) - there's no path to
0, Enter, Q, or Esc from a gamepad. Documents the real Esc-to-exit
behavior and full FCTN key row (Alt+1 through Alt+9, Alt+=), both
confirmed against the emulator's own bundled doc/README.html.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
ti99sim-sdl supports --fullscreen directly (confirmed via its own
--help output) — add it to the launch command so TI-99/4A games open
fullscreen the same way every other standalone emulator here does,
instead of a small windowed default.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
The README's "Return to launcher: Ctrl+Alt+Shift+W or START+UP+RB" was
wrong on both counts — confirmed live the controller combo did nothing.
The real, standard Moonlight client shortcuts (confirmed against
Moonlight's own documented shortcuts, not Wolf-specific) are
Ctrl+Alt+Shift+Q on a keyboard and Back/Select+Start+LB+RB pressed
together on a controller. Fixed in all three places this text appeared.
Also documented the actual working procedure for configuring a Cemu
controller and TV audio device, worked out live this session: both need
to be done from the Desktop (XFCE) app rather than ES-DE, since Cemu's
own Settings dialogs don't render correctly in ES-DE's Sway kiosk
session but do in a real XFCE window manager. Desktop shares the same
~/.config/Cemu as ES-DE, so the config carries over. Also notes the
real fix for Cemu's TV-audio-device hang (a full Wolf restart, not just
reconnecting) that worked in testing, and that /ROMs is now mounted in
Desktop too (previous commit) so Cemu can load games directly from
there.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
The 'desktop' catalog entry only ever mounted emulators/ and the
persistent home — /ROMs didn't exist in that container at all.
Confirmed live: with the controller fix requiring a real XFCE session
to reach Cemu's own working dialogs, there was no way to actually load
a ROM once there (Cemu's own File/Load menu had nothing to browse to).
Add the same /ROMs and saves mounts es-de/retroarch already use, so a
standalone emulator run from Desktop can load games directly.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
install_wolf()'s own copy of the app-injection step used $APP_KEYS instead
of its actual variable $_APP_KEYS (the underscore-prefixed name is
install_wolf()'s own; APP_KEYS is manage.sh's separate copy's variable).
Confirmed live: under 'set -u' without 'set -e' this was a non-fatal
"APP_KEYS: unbound variable" error printed mid-install right after "Wolf
restarted with updated config" — the script kept going, but the
Steam-storage-location success message never printed.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
ghcr.io/games-on-whales/xfce has no libfuse2/libfuse3 at all (checked
against its own Dockerfile) and, unlike es-de's Dockerfile (which sets
this exact env var), no fallback either. Confirmed live: launching Cemu's
AppImage from a terminal inside the Desktop session failed outright with
the standard "AppImages require FUSE to run" error, even though the same
AppImage runs fine when ES-DE launches it — es-de's container already
carries this env var, xfce's doesn't. Adding it makes every AppImage
self-extract into a temp dir and run from there instead of trying to
FUSE-mount itself, matching how ES-DE already handles them.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
update_field() only ever got called for 'mounts' and 'env' when an app was
already present in config.toml, so the games-on-whales/desktop -> xfce
image-name fix from the previous commit would NOT have reached anyone who
already has a (broken) Desktop entry — re-running the installer or
./manage.sh apps would keep refreshing mounts/env but silently leave the
old, 404ing image reference in place forever.
update_field's own array-reformatting logic (scan forward for a closing
']') isn't safe to reuse for a single-line scalar field like image or
icon_png_path — there's no ']' on that line, so the scan would run into
an unrelated array further down the same block (e.g. 'ports = []') and
corrupt it. Added a separate update_scalar_field() that only ever
replaces the exact matched line, and wired it in for both 'image' and
'icon_png_path' in both copies of this app-injector script (install-time
and manage.sh's own 'apps' command) — verified locally against a
synthetic config.toml block that it replaces only the targeted app's own
fields and leaves a neighboring app's identically-named fields untouched.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
ghcr.io/games-on-whales/desktop never existed. Confirmed live: Wolf logged
"[DOCKER] error 404 - No such image: ghcr.io/games-on-whales/desktop:edge"
and silently returned to the Moonlight app list with no other visible
error, making the Desktop tile look like it just didn't launch. The
games-on-whales/gow repo's apps/ directory names this app "xfce", and
ghcr.io/games-on-whales/xfce:edge is the real, currently published image
(confirmed against GHCR's own tag list for that package) — the icon path
uses the same "xfce" naming.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
The 'desktop' Wolf app (ghcr.io/games-on-whales/desktop:edge, full XFCE
session) had mounts=[] — no access to emulators/ or the persistent
.config/.local/share home ES-DE/RetroArch use. That meant there was no way
to run a standalone emulator like Cemu outside ES-DE's single-app Sway
kiosk session, which matters because a second top-level window from an
app (e.g. Cemu's own Settings/Input dialogs) can fail to ever get mapped
or focused there — confirmed live: clicking Cemu's General Settings
highlighted the menu entry but no dialog ever appeared, with no way to
recover control short of killing ES-DE.
Give Desktop the same emulators -> ~/Applications and retro-home(-data)
-> .config/.local/share mounts esde/retroarch already use, so Cemu (or
any other standalone emulator) can be launched directly from a real
multi-window XFCE session instead, using the exact same settings.xml the
next ES-DE-launched session will read.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
ES-DE's findEmulator() decides found-vs-not-found from the <command>
string's emulator token, and every real %INJECT%=...esprefix example in
ES-DE's own shipped es_systems.xml (Dolphin/PrimeHack/Triforce/Supermodel)
pairs it with an %EMULATOR_X%/%CORE_X% placeholder, never a literal path.
The previous ti994a <command> used a literal "/bin/bash -c ..." after
%INJECT%=%BASENAME%.esprefix, which ES-DE reported as "emulator not found"
even though /bin/bash obviously exists on the container.
Fix: give ti99sim-sdl a real custom_systems/es_find_rules.xml entry
(TI99SIM, staticpath ~/Applications/ti99sim-sdl) and reference it via
%EMULATOR_TI99SIM%, matching the pattern every built-in standalone
emulator uses. %STARTDIR%=%EMUDIR% replaces the old shell "cd && ..."
prefix to keep ti99sim-sdl's working directory at its own install dir,
where it looks up the console ROM via a plain relative path.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live: the command-line fix from an earlier commit (cd into
emulators/ before launching ti99sim-sdl) never reached the user's actual
es_systems.xml, because the write step's guard was "skip entirely if a
ti994a entry already exists" — which it did, with the old pre-fix
command line still in it. The check only asked whether an entry existed,
never whether its content matched the current template.
Now always strips any existing ti994a <system> block and re-adds the
current one fresh on every run, via a small python3 rewrite (regex block
removal + re-append) instead of a blind append-once guard. Tested against
a stand-in file matching the real stale content, and for idempotency
(second run doesn't duplicate the block), before pushing.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live: RetroArch's system_directory on this Wolf setup is
"~/bioses" (GoW's own shipped default, confirmed by reading a real
retroarch.cfg directly — wolf.sh never sets this itself), not RetroArch's
usual default of ~/.config/retroarch/system. The Dolphin Sys-folder
automation added earlier this session assumed the usual default and wrote
to retroarch/system/dolphin-emu/Sys on the host — a path RetroArch's own
config never actually reads, so it would have sat there doing nothing.
Now reads the real configured system_directory value out of
retroarch.cfg instead of assuming, and self-heals by moving a Sys folder
that's already sitting in the old wrong location (from before this fix)
into the correct one. Also corrects the same wrong assumption in the
MAME samples-directory documentation, which pointed at the same
retroarch/system/ default.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
The build succeeded, but the follow-up message pointed at "ti99sim-sdl's
own README" for where the console ROM goes — useless, since that source
tree lived inside the throwaway build container and was already gone by
the time anyone could read it.
Got the real answer from RetroPie's own configure_ti99sim(): it symlinks
the console ROM into the emulator's own install directory, then cd's
there before launching (pushd "$md_inst" && ./ti99sim-sdl "$@") — so
ti99sim-sdl finds it via a plain relative lookup, not any search path.
Caught a real bug this exposed: our own ES-DE <command> line launched
ti99sim-sdl by full path with no cd, so even a correctly-placed file
would never have been found. Fixed to cd into emulators/ first, matching
RetroPie's own pattern exactly.
Also, instead of just printing the required path (emulators/TI-994A.ctg)
and leaving it at that, now prompts for it directly with a tab-completing
path picker (read -e -i, defaulting into the game storage dir) — if you
already have the file somewhere, it copies it into place immediately
instead of making you do it by hand afterward.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
You were right to push back on this earlier — the original ask was SDL2,
same as RetroPie, and I substituted a different thing (SDL1-compat) without
checking what RetroPie actually does first.
Checked RetroPie's own ti99sim.sh scriptmodule directly: it doesn't build
from billzajac/ti99sim (the GitHub fork the last two attempts used) at
all. It fetches upstream v0.16.0 straight from the original author's own
site (mrousseau.org), applies exactly one trivial patch (a missing
#include <cstring> for modern g++), and builds against libsdl2-dev +
libssl-dev directly — no SDL1 compatibility layer, because that source
genuinely supports SDL2 natively. The GitHub fork used previously was
just a stale, pre-SDL2-port copy under a different account.
Verified the download URL actually serves the real tarball (fetched real
binary content, confirmed the 7zXZ header) and the patch's exact content
before using either — same real-source-first standard as the wolf.sh
fixes earlier this session. Tested the sed insertion that replaces
applying the .diff against a stand-in file matching the patch's real
context before trusting it.
Also documents the exact BIOS-equivalent filename RetroPie's own docs
give for this emulator: TI-994A.ctg (case-sensitive) — a real, useful
detail that was missing before.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Confirmed live: the previous sdl-config shim got past the tooling
problem, but hit a much deeper one — ti99sim's SDL frontend
(src/sdl/main.cpp) is written against the actual SDL 1.2 API and uses
symbols SDL2 genuinely removed (SDL_keysym, SDL_WM_SetCaption,
SDL_EnableUNICODE, SDL_JoystickOpened, the old integer-index
SDL_JoystickName). No shim can paper over missing symbols — this needed
the real thing.
Swapped libsdl2-dev + the hand-rolled sdl-config shim for
libsdl1.2-compat-dev — the SDL project's own official compatibility
package (libsdl-org/sdl12-compat, packaged in Ubuntu's universe repo):
genuine SDL 1.2 headers and its own real sdl-config, implemented on top
of actual SDL2 underneath. The code compiles against the API it was
actually written for, while still running on modern SDL2 at runtime.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VLX1yYKJExGSXmgUhxKQG6
Outis