Since 7f20702b4 an omission read in an emulator's own folder is one per
folder, and the folder is the core's: mame2015 and mame2016 each list
3dobios.zip. The check keyed omissions by root and dest alone and read
seven manifests as listing one place twice.
RetroBat installs its stable archives, so each config_reads entry is
read at the build the binary names: Hatari 2.4.1, hatariB 0.3,
Hatari2014 at ab55c3e from the buildbot, Kronos 2.7.0 and Ymir 0.3.3.
SSF ships PreviewVer R28 and YabaSanshiro 1.20.32 has no public
source: binary_sha256 names the file their address refs are read in.
The seven RetroBat records citing an option list now hold.
A launcher record citing a config, option or feature file binds only
when the platform's code installs the file (installs), its launch code
was read for what replaces it (overrides), and the emulator's profile
shows where the shipped build loads it and reads each key
(config_reads, read at pin_build). Unproven records place no file and
hand no directory. verify, cross_reference and the platform page list
them apart, outside severity and --strict.
A source_ref part that names a compiled image now has to carry an
address, a decompiled function or an IL method. An import directory, a
string table or readelf output only says the binary exists. A name
under a directory is a file of the tree and a .dll.map is not a DLL.
The exception list is empty: every closed profile now cites code.
validate_schemas now refuses a files[] source_ref whose parts are all data or build files (a table, a makefile, meson.build, the file itself). A ref passes when one part names a source file or a binary with an address. The closed-source profiles whose refs are still prose about the shipped binary are listed in DATA_ONLY_REF_EXCEPTIONS with a reason each.
validation: [romdir] walks a PS2 ROM directory the way the profile's
fork reads it, masked_sha1 hashes a file with a byte range cleared, and
size_multiple bounds a size to a granule. The resolver name steps, the
pack and manifest scans, the emulator expansion and verify call the same
content_fits. A PCSX2 companion is taken only beside an image that
passes, and a companion placeholder no longer seeds a scan of its own.
Since ddafc2ff4 a pack's manifest.json holds install_list and
targets, which install.py reads with the pack alone. The pack schema
refused both. It now names them, and validate_schemas.py checks each
against the install and target manifest schemas.
Three changes to the search-list model, one commit because the readers
and the profiles must move together for the schema to hold.
Set requirement in a groups map. pom2 falls back on apple2.rom for the
][ and the ][+, and on apple2e.rom for the //e it boots and for the
unenhanced //e: one required on the file could not say both. A profile
now declares groups: {name: {required}}, a member writes no required
of its own (schema and validate_schemas refuse one) and load_emulator_
profiles derives it, true when one of its sets is. group takes a name,
a list, or a map from set to rank where the rank differs per set. pom2
is grouped from SystemProfile.cpp and MainWindow.cpp:246-296 at its pin.
One member per required set under --required-only. The pick is the
first the code reads among the held members the region filter leaves,
by rank then declaration order; a set the platform declares a member
of is carried by that declaration. The emulator pack, verify
--emulator, the platform pack, its manifest and verify --platform read
alternatives.required_only_spares or core_file_spares.
variant_group is now selection_slot: files a core chooses between by
a rule other than first found (disc territory for PicoDrive and
Pizza Boy SC, the o2em_bios option, a WebAssembly probe in SWF
Player). Where the code takes any one, the field became a group:
ARMSX1 (ArmSxBackend.java:149-157, name order), ePSXe (load_bios
0x433540-0x433621 of the 2.0.5 Linux x64 binary, one configured path,
HLE when BiosHLE is set), FPse and FPse-NG, DXX-Rebirth's HOGs
(inferno.cpp:472-512). odin and zesarux carried it with no effect.
slot.py compares ranks only between candidates that share a set.
A core that loads the first of several names (yabause: saturn_bios.bin,
sega_101.bin, mpr-17933.bin) needs one of them. Profiles wrote this as
required on the first name and optional on the others, so verify and
the gap report called saturn_bios.bin a missing required file while
sega_101.bin was held, and a set held only through a name marked
optional read as a need met by an optional file.
Model: an explicit group on the file entry, mirroring the platform
field. priority was not reused: it is a rank, read by slot.py, region.py
and truth.py inside a (system, region) slot, and that slot is not the
set. px68k-sdl ranks its IPL list and its font list each from 1 in one
system, pom2 ranks one list per machine model, DuckStation ranks 105
images across every region and takes any of them. A rank cannot say
where one list ends. priority stays the order inside a group.
Readers, through alternatives.settle_sets:
- verify --emulator and the emulator pack: a held member fills the
set, the others leave the report, a set nobody holds is one gap
under its first member naming the others.
- find_undeclared_files (verify platform, core complement of packs and
manifests): same rule, a platform declaration or launcher counts as
answering the set.
- cross_reference: excused members are set aside as other versions.
- slot.py: ranks compare members of one set only.
- archive members: a required set inside an archive is met by one.
- truth: a member gives no requirement of its own, one_of and
one_of_required carry the set, and the exporters and the diff leave
such a line's required as the platform wrote it.
- validate_schemas: members of a group share required.
RetroPie was the only platform with status: archived, which kept it out
of verify --all, slots, the pipeline and list_platforms unless
--include-archived was typed. release_pack already decides which packs a
release publishes and the installer serves every platform, so the status
set nothing apart. The field, its schema entry and the flag are gone, and
RetroPie is covered by every full run like any other platform.
Without jsonschema the documented command stopped on an ImportError
traceback. Run as a command it now prints the install pipeline.py names
for the same step, with the format-nongpl extra, and exits 1. An import
from the tests still raises ImportError, which they skip on.
campaign.py takes buildbot nightlies (or a build of the pinned commit), the test content of tests/campaign/content.yml through romfetch, and runs core_probe for the files, boot, path, alias, variants and bench checks. A fixed rule table settles hle_fallback, required, path and aliases with the record ids as evidence, the rest goes to measurements/findings.jsonl. Resumable state, load, disk and budget guards, retries, content reference counting with an LRU cap, and an autorun job after the nightly.
Each line of measurements/<emulator>.jsonl now holds the whole run:
every repetition, core logs, per-frame hashes and raw timings. The gzip
data files, the data schema and the split and join code are gone, and
core_probe --record appends one line. record_version moves to 3.
The seven records are rewritten once: each new line equals its old line
joined with its data file, and rebuilds the first-layout file its id
hashes.
The site compiles the index and the SQLite files from measurements/
instead of a pinned clone. The pin, the sync script, the CI clone steps,
the freshness area and its refresh handler are gone. core_probe --record
writes into measurements/ of the checkout by default, and a test fails
when a line or data file HEAD holds changed or vanished.
measurements-pin.json names the measurements repository and the commit read. sync_measurements.py reads the clone in .cache/measurements at that commit, checks every line against its id and sidecar, and writes the site index and the SQLite databases. --fetch refuses a remote that changed or removed a line the pin holds, --bump moves the pin. validate_schemas checks every line and sidecar against their schemas, check_freshness flags a pin behind the remote head.
core_probe.py loads a libretro core with no window, runs it for a
number of frames and reports the files it opened (strace), its log
lines, frame, audio and save-state hashes and its speed. bench repeats
a run after warm-up and gives median, mean, deviation and range; check
swaps one system file between runs and compares one observable.
--record writes a JSON record with the core build, frontend commit,
host and every input by hash, validated against
core-probe-record.schema.json. A busy host is refused, and a
repetition that disagrees with the first is flagged.
validate_schemas.py read install-events.schema.json from the real tree while a test pointed ROOT at a scratch one, and test_schema_semantic_gate.py called unittest.main() before its last class.
--declared-only keeps the entries the platform's own list declares
(cores: null), each slot settled to the platform's copy with no core
alternative left behind. Install, --check, --dry-run, --events,
--list-systems, --system, --region, --emulator, --target,
--installed-cores, --manifest and --from-pack apply it. --core,
--list-cores, --list-emulators, --list-platforms, --list-targets,
--standalone-copies, --restore and --tui refuse it, and the interface
does not open by itself with it. On Batocera it gives 353 files of
14531, the list batocera-systems checks.
Event decision: no schema bump. A `declared` key on `file` would add a
key to an existing kind, which schema 1 forbids. A new kind, `declared`,
follows `check` with the same counts over the platform's own files and
their dests in file-event order. Readers already skip a kind they do
not know (rbevents.py and the interface reader were checked), so
EVENT_SCHEMA stays 1.
An unknown --system is answered with the systems the list files its
key file under: --platform retroarch --system snk-neogeo now suggests
arcade, which holds neogeo.zip, instead of snk-neogeo-cd. neogeo, mvs
and aes follow the same route. Never taken, only suggested. A system
title the manifest carries (SNK Neo Geo CD) is taken for its id when
no other system shares it and no platform name reads the same.
schemas/install-events.schema.json is written from EVENT_KEYS by
scripts/generate_event_schema.py. validate_schemas.py reports a stale
file, and tests compare it with EVENT_KEYS and validate a real run.
Bootstrap pins recomputed.
release_pack: true in the registry is the list a release builds: RetroArch,
EmuDeck, RetroBat, Batocera, RomM and MiSTer, 83% of the downloads of
v2026.10.10. generate_pack --all, --verify-packs, the pipeline consistency
step, release_record and check_release read it. Manifests are still written
for every platform.
batocera-systems lists bios/saves/dolphin-emu/Triforce/segaboot.gcm, so the pack folder for the save directory is userdata-saves/. validate_schemas refuses a pack folder the BIOS tree already uses.
Batocera, KNULLI, Recalbox, RetroBat and RetroDECK hand a libretro core <saves>/<system> at each launch. Each copy lands below every system that runs the core. EmuDeck keeps one unsorted saves/retroarch/saves. The unsupported platforms record why.
load_from: save_dir entries of a libretro build were left out of every
pack. They now go below the platform's save folder (registry
save_directory: RetroArch saves/, Lakka savefiles/, by core), the
manifest marks them root: save, and verify replays the same placement.
Platforms keeping saves per system or per game list them as omitted.
generate_site.py writes the Starlight project in website/ and nothing
else: mkdocs.yml, the MkDocs hooks, theme overrides and stylesheet go,
and the wiki is written in Starlight's syntax. Both workflows set up
Node with the official actions/setup-node pinned by commit, run npm ci
from the committed lockfile, build with the strict link validator and
check the result with validate_site.py.
The search weighs a page by what it answers: Start here pages first,
a system page's cards and file names before a guide citing them, deep
guides, launchers, aliases and frozen copies after. Status badges,
hash blocks and the export change lists leave the index. The ranking
sits at the centre of a range where the 13 queries of search_gate.py
all answer, in the search box itself.
839 pages, every URL and 45,611 ids match the MkDocs build. The tests
read the Starlight content and syntax.
Hashes, link lists, install commands, member tables and long reference
tables leave the site search: the index drops from 21.0 MB (4.88 MB
gzip) to 6.8 MB (1.33 MB gzip). Names and prose stay searchable.
files/?hash= takes a SHA1, MD5, CRC32 or SHA256 and reads one shard of
api/v1/lookup/ (about 22 KB gzip, was 675 KB) listing the collection
card, the platforms that declare the value and the profiles that check
it. files/?sha1= still opens the card, and a sha1- anchor missing from
a page is looked up there, so a card can move between pages.
Excluded blocks are section elements: the search plugin compares
elements by tag, so a div ended at the first nested div.
The semantic check read systems: only. mame2010 filed seven BIOS root
sets under systems its list lacked: cpzn1, cpzn2 (zn.c:4543, 4563),
konamigv (konamigv.c:877), skns (suprnova.c:1686), alg_bios (alg.c:782),
maxaflex (maxaflex.c:541) and hng64 (hng64.c:2189) are GAME_IS_BIOS_ROOT
in mame2010-libretro at source_commit.
The -listxml names, for every machine, its parent (romof), the devices of its default configuration (device_ref) and the cards its slots offer (slotoption). A machine takes the system of the first set on its romof chain a written entry lists, and passes it to each device it can instantiate. A device set reached from one system gets system:, from several the new per-entry systems:, from none neither, and keeps serving every system of the profile. The gap report, the core extras and slots read the entry's own systems, so generate_pack --system and the installer pick by them.
An emulator page linked its YAML and api/v1 JSON, platform and system
pages nothing. Each platform page now links its YAML, its own
api/v1/platforms/<id>.json and its install manifest, and each system
page its api/v1/systems/<slug>.json, both validated by the envelope
schema.
Profiles, the wiki, comments, docstrings, command output and schema descriptions carried em-dashes, clause-joining semicolons, unicode arrows, box-drawing banners and check-mark glyphs. Each sentence was rewritten rather than having its character swapped. Hashes, paths, source_ref values and every cited file:line are unchanged, which a comparison of each parsed profile against the previous revision confirms.
validate_pr comments now say PASS and FAIL in words, and source_ref values in a profile's structured blocks render as code. test_prose_style fails if any of these comes back, in tracked files or in the generated site.
argparse completed a prefix into the only flag starting with it:
slots.py --emulator read profiles from a directory named after the
emulator and reported nothing, and pipeline.py --all built every pack
variant. Every parser now refuses an abbreviation.
A comment named the Bluetooth directory twice, a path no revision
holds, and every firmware ref sat ten lines above the case it meant.
Each now cites its case and return at the pinned commit. The schema
check reads profiles as text and refuses a citation that repeats
its own directories.
kenji-nx cited tmp/es-de/ANDROID.md:470-474, a path from the machine of
whoever profiled it. No revision of any declared repository holds it, so
profile_sync could only report it missing, every pass, forever, and no
amount of reading would ever settle it.
validate_schemas now refuses a scratch directory, an absolute path, a
Windows drive path and one climbing out of the tree, and names the
offending citation rather than the scalar that carries it. Offline, so it
runs on every push and every pull request rather than waiting for a
network pass.
The ES-DE citation reads as external now, which is what it always was.
kenji-nx is at 37 refs, all anchored: the three changed blocks were var
giving way to explicit types.
install.py fetches a file from its repo_path or from a release asset.
Resolution can land on a file the database does not index, and the
entry then shipped with neither: a line in the download list that can
only ever fail. Those are recorded as omitted instead, which is what
the installer already knows how to report, and a test holds the
committed manifests to it.
validate_schemas read dist/ while a build was writing it and reported
a half-written pack as 'File is not a zip file'. It takes the shared
lock --verify-packs uses, and says so when a build holds it.
Two entries claimed bios/Sony/PlayStation 3/PS3UPDAT.PUP with different
SHA1s. Preserving large-file entries matched on path and keyed on
SHA1, so replacing a firmware revision on disk left the old entry
pointing at a path that now serves other bytes. A preserved entry whose
path the scan has already claimed is dropped, and validate_schemas
refuses a database where one path carries two entries.
Separately, a run without --force rebuilt each cached entry from a
hand-written list of four digests and wrote it back without adler32,
so one such run stripped the digest from every file permanently. A
cache entry missing any digest is now a miss.
JSON Schemas for the database, install and pack manifests, target
manifests, site API envelopes and stats, plus the semantic invariants a
schema cannot express: declared totals matching their lists, no
destination both installed and omitted, database keys matching their
sha1. validate_site.py checks the rendered HTML for metadata, headings,
image alternatives, duplicate ids and unresolved local links.
Pack manifests are read from inside the generated archives, where
generate_pack writes them, rather than from a dist/ glob that matches
nothing.
Emulator and platform schemas gain additionalProperties: false, and
cores[] plus contents[].name must be strings: an unquoted 81 or 01 in
YAML parses as a number and stops matching the upstream name.
Abdessamad Derraz
dajiaohuang