The search splits a page at its headings. Emulator cards were titled in
bold and platform systems folded into blocks under one heading, so a
pasted hash opened the top of a page or the systems overview of a
100-block platform page. Each file card is now a level 4 heading kept
out of the table of contents, and each platform system a heading.
Card lines that list other pages (platforms, packs, source citations,
collection paths) leave the search index: 24.7 MB down to 18.1 MB, the
names, hashes and notes still searchable.
The 3.12 guard rested on generate_readme f-strings that have since been
rewritten. Every script compiles under CPython 3.9.25, 3.10 and 3.11, and
only generate_pack's parenthesized context managers fail under 3.8. Under
3.9 the offline pipeline (--skip-packs) completed with no change to the
committed artifacts, the 15 manifests matched 3.12 byte for byte, and the
RomM pack was identical. CI now compiles every script under a real 3.9.
The recipe snapshot kept only the archives a profile or platform names,
so provenance_report --missing named no set for dumps a public romset
carries (fm7740sx holds the FM77AV40SX DICROM, EXTSUB and KANJI2 ROMs).
The importer now also keeps the sets holding a catalog entry the
collection lacks. Reimported from mame0289 and the cached FBNeo DATs:
MAME 1726 to 2148 recipes, FBNeo 104 to 111, none dropped.
The help said "Only regenerate DB and verify" while steps 1b, 4d, 7 and
8 still ran and rewrote derivable.json, README.md and docs/. It now names
the steps it skips and points to --skip-docs.
The triage, --changed-only and --strict counted refs only. Prose that
--realign-prose would move or refuse blocks every recale and bump, and
blastem read 0 to review. Each report now carries that reading. The
triage also gets a header, a closing count, and one sampling notice per
run instead of one per profile.
architecture.md drew path_suffix ahead of SHA1 and listed a core
resolution that missed the retroarch list entry and built_by. Both now
link to the one statement that matches common.py. The Recalbox table
read hashMatchMandatory into verify.py, which never reads it, and
md5_composite hashes member contents, not member MD5s
(Zip.cpp:71-96). The validation index is built from the platform's
resolved profiles only. The requirement table gains the fields that
make a platform YAML reversible, and pages that name a platform list
say what changes rather than judging it.
The installer output example showed 5865 files and no Not installable
line, while the RetroArch manifest gives 8069 and omits 11. Exit code 2
also comes from every option argparse refuses. RETROBIOS_REF holds a
file list only until the next release deletes its tag, and large files
always come from large-files. The release commit now carries the
manifests the pipeline regenerates. A RetroPie 4.8 image ships Python
3.7.3, below the bootstrap's floor, and the pages say what to use.
The test skeleton indexed verify_platform's dict result as a list and
stopped on KeyError: 0. The verification discipline and the profiling
guide hedged rules the project holds as firm. The profile field table
marked as required fields the schema does not require, and now names
the pin that profile_sync --strict needs.
A value of another length was looked up as an MD5 and reported NOT FOUND,
a SHA-256 was dropped, and an input with no hash ended on "No valid
hashes found in input". Digests are now read by length (CRC32, MD5 and
its 29-digit Batocera form, SHA-1, SHA-256), SHA-256 is looked up in the
database, and any other value is refused with the formats accepted.
kronos tags the ST-V BIOS epr-19854.ic8 taiwan, mednafen_stv tags it
asia-ntsc, the SMPC area Mednafen names for Taiwan (ss.cpp:1119,
2363-2364). As siblings under asia they never matched, so a request in
one spelling dropped the file tagged in the other, and --region
taiwan,us dropped the Saturn sega_101.bin. AREA_TERRITORIES relates
each area to the territory it serves, china for asia-pal included, in
region.py and its installer copy.
The coreinfo area compared .info names only: a firmwareN_path added to a
profiled core's .info came out OK. check_freshness now compares each
profiled .info's git blob with platforms/_core_info.json, recorded on
2026-10-21 from the current listing, and --record-coreinfo CORE records
one as reviewed.
mkdocs was installed by the deploy alone, after the merge, and wiki/
was not a validation path: a dead link or a stray heading failed the
deploy on main with nothing in front of it. The test job now runs the
deploy's site build, and every site input triggers validation.
The Required packs and manifests carried every data tree, the FreeIntv
overlays included. A data_directories entry now accepts required: false,
and the platform, emulator and manifest walks leave such a tree out under
--required-only. FreeIntv reads its overlays only with the multi-screen
option and draws a placeholder keypad without one (src/libretro.c:294-296,
350-369).
The cache was truncated and dumped in place. A concurrent build read
half a file, took it for empty, and an offline run then wrote {} over
every recorded icon. The file is now replaced through a scratch file,
merged with what another run recorded meanwhile, and left alone when
nothing new was learned.
--verify-packs after an --all-variants build rebuilt only the full pack's
name and opened it alone. The five narrowed packs of the same run passed
without a line. Each one is now named with the hash check the build ran
on its members.
Names are matched on letters and digits, so the site's display names
work, and a typo gets the closest names instead of a full list. Hints
and advice give back the command the user ran (curl or PowerShell form,
every flag kept), --dest alone follows detection and says whose list it
installs, Ctrl+C at a question exits 130, and --cores-dir names a core
folder. install.ps1 runs in its own scope so irm | iex leaves the
session alone, and an override that causes a fingerprint mismatch is
named. --check lists hash mismatches first, missing required files are
grouped by system, and --help ends with examples.
--emulator accepts a buildbot core name from a profile's cores:, and a
mistyped target gets the closest names instead of every target. diff_truth
shows destinations and reports unpaired entries as evidence, not removals.
The verify summary counts discrepancies, a shared destination names both
systems, a truth hash conflict names both cores, and pipeline steps carry
their place in the real run instead of N/8. Every GitHub API caller takes
the token from github_token(), which falls back to gh auth token.
Troubleshooting, the FAQ and the download page sent people who only ran
the installer to verify.py, which reads the repository, not their BIOS
folder. Download links now name the exact asset of the recorded release,
the --dest example names the BIOS folder on the card, the slice layout
left with the last release that used it, and the shown verify output is
what the tool prints.
A fork's GITHUB_TOKEN is read-only, so the comment never arrived and a
valid file turned red. The report now goes to the job summary for every
pull request, and the comment and labels run only where the token can
write. validate_pr points an oversized file to an issue instead of a
platform setting a contributor cannot use.
Principles, collection rules and the inverted model in architecture.md, research and citation rules in profiling.md, platform and scraper rules in their pages, tool conventions in tools.md, test and commit expectations in testing-guide.md, installer constraints in installer.md. Each rule sits where a reader looks for it, with its reason. Also fixes the stale release-job line, the suite command without -t ., and the unsourceable definition.
On Windows the partial copy of a file of 16 MB or more is locked with msvcrt.locking, a LockFile region: the locking process cannot access that region through a second handle (Microsoft LockFile remarks). Hashing reopened the path, the read failed with a sharing violation, and every large file was reported as 'folder not writable' (issue 82). Both digests are now read in one pass through the handle that holds the lock. A read-only file already at the destination had the same misleading report: on Windows its attribute is now cleared before the verified copy replaces it, and a refusal that remains is named for what it is.
RetroPie's YAML listed no standalone_cores, so every package was laid
out as a libretro core and the standalone files of amiberry, dolphin,
hatari, mame and ppsspp never reached the pack. Packages without the
lr- prefix (bar the retroarch frontend) are now standalone, and names
shipped both ways are dual-mode.
A names_searched list is one place the launcher fills with the first name it finds, so a core file it would never open once another name of the list is held is not reported missing. A place no declared file and no packed core file fills receives, under the launcher's name, a file the core accepts by content. The layer is applied inside find_undeclared_files, which verify and the pack builder both read, and a test reads their source to keep it there.
The scraper walks each package's functions the way RetroPie runs them in one shell (assignments, arrays, for loops, brace expansion, $md_id, calls into other packages) and records every link, copy, move and setting that names $biosdir, the addEmulator flags (sdltrs -romfile3), the BIOS read outside $biosdir (basilisk --rom, gngeopi -B) and the launch scripts that exit without their file (fs-uae kick13.rom, reicast dc_boot.bin).
Each package also gets a package record with its rp_module_repo and the systems it registers, which maps a package to the build it compiles. The schema gains package, repo, systems and the moves and package kinds. platforms/retropie.yml is regenerated at 46128cd2: only the launchers section is new.
The scraper now records, from configgen at the pinned tag, the names each Recalbox launcher tries and where it hands them: the Hatari TOS cascades (standalone, hatari and hatarib cores), GSplus, Amiberry kickstarts, the PCSX2 region choice, the Dolphin IPL presence test, xemu, the mupen64plus 64DD IPL, the Mini vMac boot disk and the o2em_bios option. The system_directory libretroRetroarch gives each core, and the per-system overrides after it, are passes_path records.
The scraper reads the start scripts and the configs each package installs at the tag the BIOS list is pinned to: configs that name a BIOS path, the yabasanshiro search, directories handed to DuckStation and the PCSX2 builds, and the ln -s commands that put bios/ where an emulator reads. A package that ships one config per device gets a device field only where the values differ, as melonDS does on AMD64.
The scraper parses the configgen generators with ast at the pinned tag: DuckStation and its legacy build, Hatari, xemu, melonDS, mupen64plus 64DD, x16emu, GSplus, and the o2em and neocd core options. An extractor that no longer recognises its code raises instead of writing an empty list.
The page named the entries a change touched and never the values, so
a maintainer could not check a correction. The report now carries
changes, one object per change with kind, entry, field, old and new,
read from the platform entry and from the value the render writes
(written() per format: record value for System.dat and RomM, one md5
where a single one is checked). The existing label lists are derived
from it. The page writes "md5 X becomes Y" and names each required
flag with its new value. A test parses the original and updated
System.dat and es_bios.xml and checks the list equals the diff.
The export keeps every entry a platform declares. An entry the truth
does not pair was still read in 1189 of 2776 cases (alias, hash, a
type: directory tree, a core outside the resolved set), so absence of
a pairing proves nothing. Each report now lists, under
removal_candidates, the unpaired entries that no profile of the cores
the platform runs answers by name, path tail, hash, directory or data
directory, with the cores examined and a reason. A core the platform
names without a profile, or a ROM set under a core that opens sets by
name, leaves the verdict undetermined. The exports page states the
count and lists the entries per platform.
The resolver reads built_by as the platforms on which a core name means this build. Its definition said platforms that compile a libretro build, which excludes a standalone build a platform installs from the project's release. The field now says what it decides: platforms that ship this build under its core name, compiled or installed from the release.
DuckStation identifies its BIOS by MD5 in a folder it scans
(bios.cpp:208), so its entries' names are labels. The export wrote them
into Batocera's and ROCKNIX's lists as files to hold: a user with only
scph5501.bin read 104 missing BIOS where DuckStation boots. Labels also
lent their md5 to a name another core opens, and paired by name they
replaced System.dat's scph5000.bin hashes.
found_by: name | content, on a profile or a file and keyed by build like
load_from. The truth records which cores open each name (_named_by). A
label pairs with a platform entry by hash only, is never added, merges
with an opened name only on a shared hash, and is counted in the export
report rather than listed as missing.
A machine that instantiates a device loads the device's ROMs from <shortname>.zip (src/emu/device.cpp:140-152), and a missing required ROM stops that machine (romload.cpp:529-570, 667-671). The BIOS root scan never sees these sets. mame_hash_scraper --devices --listxml reads them from the -listxml of the release mame.yml pins, refuses another release, and cites each set at the ROM_START block holding its ROMs. The BIOS root merge also wrote refs without the upstream prefix and would have rewritten all 171 of them in mame.yml.
Profile notes and wiki pages called .info entries phantom or incorrect, and labelled an upstream list a known defect. They now state what each file declares and what the code reads. The wiki claimed Batocera prefix-matches truncated md5s; its script compares the full string (batocera-systems:1088), which the page now says. The gaps column reads "Not read by profiles" and notes that a profile may be incomplete.
The page called the original "their file" and listed hashes "corrected" and additions "left out", a verdict on work the page only proposes to extend. It now says what each updated file changes in neutral terms built from the report's numbers, gives the reason a file is not added, credits the effort a list takes to keep current, and invites an issue when a change does not fit a platform. A test keeps the judging words off the page.
Abdessamad Derraz