A long checkout path wraps "BIOS folder:" after the label, and the
test joined the lines without the space it dropped. It now compares
without spaces, which holds for either wrap point.
Confirm has no Install row until its dry run ends, so an Enter sent
on its title alone fell on "Counting the files" and did nothing on
the runner. The test now waits for "Counted again for these choices",
drawn once the Install row exists.
Run from the retrobios package, a hint and --help named python3 install.py, a file that user does not have. The package's launcher sets RETROBIOS_BOOTSTRAP=pypi, and the command is read from the environment it runs in: uvx retrobios from uv's cache, pipx run retrobios from pipx's, retrobios once installed.
- The installer speaks without "you": the first screen asks "Where
to install?", rows read "Copy the BIOS files from an old card or a
folder", a close match reads "Close matches: ...". The handheld
launchers and their port text follow, and the pronoun guard no
longer lists any installer message.
- A setup that runs as an app opens on its one-line command, the
folder table after it; a setup that runs on its own device keeps
the network folder first, which its ZIP route needs.
- The home steps say what to do: "Its page gives one line to copy",
"Paste it, press Enter".
A folder under the checkout is as long as the checkout's own path. From a deep clone the screen broke it after a separator, as it should, and eight tests looked for it on one line.
The game-data tap test waited on the bare RetroBIOS: title, which the
screen before Confirm already showed, then pressed Enter only if one
settle had drawn Confirm. The runner drew it later and the install
never started. It now waits for the Confirm title and presses Enter.
llms.txt now gives the package commands, the plan as JSON events from
--dry-run --events, the exit codes and the plan, consent, act pattern
with --yes. Tests hold each fact to install.py and to the exit code table
of the installer page.
Every archive has a recipe in recipes/archives.json, and no page a user
reads said so. A short reference page says what a recipe is, how to
rebuild one archive and what exact and canonical mean. Its counts and the
README line are counted from the recipe file at each build.
The README gains a row of four badges: PyPI version, Python versions,
the Validation workflow on main and the license. Tests hold the badges,
the counts and the page's command.
The events transcript on the installer page is a --check run that finds
files missing. Since November such a run ends on 3, as the Exit status
table says, and the sample still showed 1.
Dependabot moved uv in pyproject.toml alone, while validate.yml and
CONTRIBUTING kept 0.12.9. Both pins move together, and a test now
fails when a package pinned in both carries two versions.
build_pypi.py stages install.py, the launcher, __main__.py, README and
LICENSE, then runs uv build with flit_core 3.12.0 pinned and
SOURCE_DATE_EPOCH set from install.py's commit. Adds python -m retrobios,
the PyPA classifiers, keywords and project URLs. CI installs uv.
Search Console reported emulators/vba_m/ as not found: an alias
profile has no page, and only the names in a profile's cores: list
were redirected. Each of the 21 alias profiles now redirects to the
page of the profile it names.
- The installer says "Every file on the list is in place." where it
said "Everything is in place.": a check covers the files the
collection holds, and the required ones it lacks follow.
- The FAQ, first install, legal and home pages state only what the
code shows (PlayStation 1 and Saturn run without a BIOS on some
emulators, save files are never touched, records are added).
- MinUI and MyMinUI show their own boot logos, and the Android,
Linux handheld and iPhone tiles show the Android robot, Tux and a
plain device icon, each credited in sources.json.
- Each Markdown copy, llms.txt and llms-full.txt name the commit
they were built from, version.json the current one, and the
answering rules tell assistants to fetch a stale copy again.
- The deploy lists the pages whose words changed (digests.json) and
posts them to IndexNow once the site is live.
Each claim in the README and the wiki now matches the code or data it
describes: line citations counted as such (93%), closed-source profiles
named, severity tied to the check mode, installs limited to files the
collection holds, and the demo labelled as a demonstration card.
make_demo.py records a second variant: the full-screen interface at the
640x480 screen of an R36S, on an ArkOS root made with bwrap, started as
the Tools menu launcher starts it (RETROBIOS_PAD=gptokeyb), driven by the
keys gptokeyb sends for the buttons. demo.json holds one entry per
variant with its alt text. No Windows variant: under pwsh on Linux the
installer's summary gives --dest /mnt/e/bios back.
The quick install box of Get your files gets an "On the device" tab with
the handheld recording, the terminal recording under the others but
Windows, and, under the text of each computer tab, a drawing of opening
the terminal and pasting (sitesteps.py, inline SVG in the theme's
colours, title and desc). The terminal steps name the physical action.
The README ran 211 lines and 2,000 words: five commands with options,
an 18-row download table, coverage figures and the build commands, all
before the help. It now says what the project does, gives the line that
installs for Windows and for the rest, shows it, lists the six ZIPs and
the setups that install by command, and leads to the site, where the
figures and every option live. The build commands sit in a closed block.
No personal pronoun is left in its text.
The SHA-256 pin already refuses any changed byte: the cap in install.sh
and install.ps1 only stops a sign-in portal's page before it is hashed,
so it can be generous. The size test now fails once install.py passes
3 MiB, so its growth is looked at, and checks both bootstraps accept the
same 8 MiB. The separate size test of the earlier commit is folded into
the one test_install.py already had.
scripts/build_pypi.py writes a wheel and an sdist from install.py, the
copy install.sh pins, and packaging/pypi. The version is the date of the
release in release.json. validate.yml now runs on packaging/ changes.
Each journey of the full screen runs on a pty against a loopback list
and asserts its key count: a first install (2 to start, 3 in all), a
muOS card holding a game (2, 3, its games read since the fix), an update
(2, 3), one system then the same again (14, 15, then 2, 3 for the whole
list), what is missing (5, 6), a folder for another device (19, 20).
install.sh and install.ps1 refuse an installer above 2 MiB. The test
fails once install.py passes 95 % of the cap both bootstraps read, so
the cap is raised or the installer made smaller before a release stops
every one-liner. Raising the cap raises the alert with it.
Choosing GBA alone brought FBNeo's cheats and samples, 3,472 files,
because both trees listed every FBNeo system. The profile now names the
systems each serves, from the code: a driver's <name>.ini, cheat.dat,
cheatnes.dat or cheatsnes.dat (conc.cpp:815-828, 1059-1083), the md_,
pce_, tg_ and sgx_ drivers the tree holds, and samples read by arcade
drivers alone (samples.cpp:886-899).
The list now gives a tree the platform's ids its systems normalize to,
as a core extra's: the cheats said sega-megadrive where RetroArch says
sega-mega-drive, and a tree no platform system serves goes with the
files of none, as the pack puts it. The installer keeps a tree for the
systems chosen by name, not for every system a shared file names: the
arcade choice took blueMSX's machines through msx.zip.
The page said a long list opens on None. The list opens on its first
system since the three tick lists share one rule (A on None unticked
every system as a beginner's first press). The page now says what the
screen does, and a test reads both.
With systems ticked since the check, the headline counted them (About
3479 files, 82.2 MB) and the Install row kept the whole list's size
(Install 6.6 GB). _ticked_estimate is the one count both read, offline
the files the folders copied from hold.
muOS names its GBA folder gameboyadvance first, which resolved to ares's
game-boy-advance. mgba lists nintendo-gba, so the fallback set was empty
and an unrecognised GBA game got no file. ares, dgen, picodrive, fbneo,
fbalpha2019 and emux_sms now use the ids the other profiles use, with
their old spellings kept as SYSTEM_ALIASES (bios.sms and fbneo's
Spectrum sets had fallen to the other bucket of the RetroArch list).
A folder whose name spells a machine its core does not run now also
holds the core's machines: a PC Engine CD folder run by mednafen_pce_fast
got no System Card. Install lists and install/games regenerated.
The full screen passes --platform=retroarch for a muOS card. Detection
never runs in that child, so games_setups fell back to RetroArch's own
setup and the run said no games folder was found beside MUOS/bios. The
card's own files now name its system, as card_app reads them.
The build writes <page>/index.md beside each page, named in its head,
and llms-full.txt. llms.txt and the setup skill say how to answer from
the site, not from memory, and the setup_bios_files WebMCP tool gives a
setup's folder and exact file names from the picker data.
ArkOS ran the GBA with gpsp_ezode_mod and dArkOS with mgba_rumble: no
profile named them, so an install for the card's games kept every GBA
core's files, FBNeo's 3,472 data files among them. Each rename proven
from the OS build (same code base, same BIOS loading) joins its
profile's cores. The resolver folds either separator, and falls back to
the emulator column when the core names a mode, a machine or a plugin
of that program, only where its profile declares the folder's system:
MAME documents no GameKing ROM. 109 unresolved names are 56, each one
recorded with why in tests/test_games_index.py.
--yes alone installed the whole list, 6.4 GB on a dArkOS card holding
two games, while the screen offers what the games need first. A run that
asks nothing now takes that first choice when the card holds games, and
says so. --all-files installs the whole list. A run given its list
(the screen's own runs) or narrowed by name is unchanged.
The manifest names the cores that read a tree no platform system
declares, every reader of a shared tree included. An install for the
games of a card keeps such a tree only when its plan runs one of them:
a SNES game on ArkOS (snes9x) no longer takes FBNeo's cheats and samples.
A system no setup names a core for keeps every core's tree.
The rerun without jsonschema took 345 s here and went past its 600 s
limit on the CI runner, about twice slower, so main stayed red. It now
selects the modules that import jsonschema or validate_schemas rather
than those that name them, and allows 1800 s.
Abdessamad Derraz