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.
It joined every setup page whole, its parts and its reference of every
file: 6.9 MB, past what an assistant reads at once. Each setup's own
page now stops before that reference, which its Markdown copy keeps.
Readers reached the 404 page with a pasted command (/i.ps1%20%7C%20iex),
a file name, a GitHub path or a section below a page: the page now says
which it is and leads on. Beside "Start with", the first file's
download sits next to its name.
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 name step reused one variable for paths and for (path, md5, size)
tuples, the standalone copies reused targets inside its own loop, and
the WSL state read as object. Ten baseline entries no longer occur.
The data tree record carries the needed flag since 072380c26, and
the manifest test still built the three-part one. profile_sync's triage
lists and _emit's returns, site_insights' GoatCounter answers and
make_demo's blocks are annotated, and the baseline loses the entry
that no longer occurs.
Main changed install.py without the launcher index, so --add-app
refused every launcher and nine tests failed on main itself. The demo
script added a list without a type, and an archive-size test still
built the data-tree refs without the needed flag the builder reads.
The README shows the GIF under the commands. The home page and Get your
files show the video through InstallDemo.astro: nothing is fetched until
it is on screen, reduced motion and Save-Data keep the poster with a play
button, and a description is read to screen readers.
scripts/make_demo.py builds an ArkOS card under tmp/demo/, mounts it in a
user mount namespace, serves the checkout on loopback and runs
docs_assets/demo/install.tape with VHS, then encodes the GIF, WebM, MP4
and poster and holds them to their size targets.
Three site commits pushed to main fail checks CI runs: two prose
semicolons, an untyped signature, and GoatCounter answers read as a
dict without a test. GoatCounter's stats now go through one reader
that refuses an answer that is not an object.
The skill and llms.txt said "a quote" where the tool refuses the
typographic ones too, which end a PowerShell string. One constant states
the set for both, and points to device_setup_command.
The skill and llms.txt told an agent to write --dest=<card>/bios bare.
A folder now goes in single quotes, attached to its option, and one that
holds a quote, a backtick or a control character, or starts with -, is
a question for the user.
ARD reserves text/markdown for agent skills, and Lighthouse warned on
the llms.txt entry. skills/setup-bios/SKILL.md holds the steps that set
up a device with the fewest files, and the manifest lists it as a skill.
llms.txt and an ARD manifest name the pages and the JSON API, and lead
an agent setting up a device to --for-games --required-only. Every page
registers four read-only WebMCP tools: the setup command, a hash lookup,
an emulator's files and the site search.
--for-games --required-only on a dArkOS card from a computer kept every
file of the list: --dest skipped the card's detection, so its rom
folders were never found. The games part beside the BIOS folder now
names them. The manifest marks a data tree every core runs without
(optional), and --required-only leaves it out: one GBA and one SNES game
took FBNeo's cheats and samples, 3,472 files no game opened.
validate.yml runs export_native with continue-on-error, and pr_checks
failed the whole run on it. The step is now shown and does not fail
the run, and a test keeps that list equal to the workflow's.
Code moved to sitecollection.py and the triage progress added errors
the baseline did not hold. A reused loop name, two lists without a
type, returns without a value, and a regex match read without a test
for None. Two baseline entries no longer occur and are removed.
og:title and twitter:title carry the search title ("mGBA BIOS
files"), an emulator page shares its maker's card, a download page is
dated by its release in the sitemap, and a Switch emulator's files are
called keys.
recipes/archives.json records each of the 2,222 archives database.json lists, .variants/ copies included: members in order with size, crc32 and sha1, the container sha1, md5 and size, exact when the TorrentZip build gives its bytes (1,790) and canonical otherwise (432), the profile and DAT recipes agreeing on its members, and the platforms or profiles pinning its bytes.
romset_recipes.py --capture writes it (pipeline step 1c), --build and --build-all rebuild from held members, loose first, then another archive, then its own copy, read back and checked before the move into place. Under zlib-ng nothing is claimed exact. A test fails when an archive has no recipe or a member is held nowhere.
The importer keeps the archive names the collection holds and every set whose member list equals an archive the collection keeps under another name (aga.zip is isa_aga), under the collection's name with the upstream one in set. All 113 MAME listxml releases (0.177 to 0.289) are imported, not one in five, and FBNeo is read again at its tip: 2,936 MAME recipes and 145 FBNeo ones, compacted as before.
fccpu20.zip, fcisio1.zip and interpod.zip now have recipes and carry slot ROMs beside their own set. Trimmed, each equals the archive under the newer MAME name, so they are kept as they are until the storage decision.
write_torrentzip reads and deflates one member at a time, gives back what it read (size, crc32, sha1) and takes a ZIP64 field only past 32 bits, so a 2 GiB disk image builds without being held whole. TorrentZip layouts are unchanged. zlib_ng_version() names the deflater that cannot reproduce TorrentZip bytes.
Recalbox hashes a ZIP by its members (zip_digest: composite, Bios.cpp:64-68), so its md5 names no container bytes: romset_recipes --missing tried to rebuild it and counted it unreproducible.
validate.yml runs mypy against its baseline and checks that shellcheck
is installed, and pr_checks ran neither. The guard test only knew the
python, vermin and npm commands, so it missed both. A step can now pipe
its output into a second command, as mypy into mypy-baseline filter.
Abdessamad Derraz