Most modules write under tmp/, which a fresh clone lacks, and only three
created it on import. A module run alone failed with FileNotFoundError on
a temporary directory.
docs/ is gitignored and was built after the tests, so the three guards
that read rendered pages skipped on every CI run. The site is generated
first, the suite runs whatever that step returns, and the guards fail
in CI when the pages are absent.
A module-level alias written with a PEP 604 union compiled under 3.9 and
raised TypeError when check_release_assets.py, and check_release.py
through it, were loaded. The floor test now imports every module under
scripts/ with a real 3.9 interpreter, which CI gives pyyaml and
jsonschema.
Thirty one-line labels in verify.py, generate_site.py and
generate_truth.py named the statement under them ("# Load registry",
"# Copy branding assets"). Comments that give a reason stay, as do the
module section headers. A test reads the three files for new ones.
profile_sync gains --changed (added or edited profiles, uncommitted
ones included) and --added, both against the merge base of --base
(origin/main by default). CONTRIBUTING runs --changed --strict instead
of one --emulator call per new profile, and the CI job runs --added
in place of its own diff loop.
FirmwareDatabase.cs:405-410 declares NDS+NAND_JPN, EUR, USA, AUS, CHN
and KOR, each with a placeholder hash and 251658240 + 64 bytes. In DSi
mode DecideNAND takes the first one set, limited to the regions a
DSi-enhanced header allows, and stops the load with none
(MelonDS.cs:114, 304, 484, 684-706). DSi mode is off by default, so the
entries are optional, like DSi_Firmware.bin and the DSi BIOS slots.
A missing extended ROM only raises a notice (memory.cpp:1833-1849), but
the machine then runs on the Kickstart alone. cdstrap, cd.device,
cdtv.device and cdfs.library are resident modules of the CD32 and CDTV
extended ROMs and absent from their Kickstarts, the CD32 and base CDTV
models have no floppy drive (cfgfile.cpp:9375-9379, 9405-9406), so no CD
is read or booted. Both entries stay required. The AROS replacement
stands in only for a missing Kickstart (memory.cpp:3055-3063), so the
CDTV entry loses hle_fallback.
FreeIntv declared no data tree while its alias freeintv_ts_overlay,
the same code, declared freeintv_overlays. The core reads
freeintv_overlays/<rom>.png only with the multi-screen option and draws
its embedded keypad image without one (src/libretro.c:278-280, 294,
321-325), so the tree is optional. A test holds every alias to the trees
of its core.
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 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.
Started outside the repository root, both ended on a FileNotFoundError
traceback: the database and the data registry name files from the root.
They now move to the root after reading the paths the caller typed, which
stay relative to where they were typed.
Every step runs from the repository root, but the lock and the purge of
stale packs read --output-dir from the caller's directory. Started from
scripts/, the run locked and purged scripts/dist/ and wrote to dist/,
leaving the stale packs there. The directory is now made absolute before
the lock, dist/ at the root by default.
Quick Install was the one Title Case heading left in the published
text. The Files sentence was glued to the paragraph above it, two
paragraphs before the table whose column it describes. It now names
the column and sits just above the table.
The lock refusal, "no pack for", "no manifest for" and "no pack to
verify" went to stdout, so a run redirected to a log exited 1 with
nothing on the terminal. A test reads both scripts and refuses an ERROR
line printed to stdout.
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.
--platform retroarch --system sony-playstaton listed the first ten systems
of the alphabet. The warning now names the closest systems and the
command that lists the platform's own, as --system alone already did.
The release-source check called any uncommitted repo_path gitignored,
pointing at the release assets instead of a missing commit, and its
skip outside a checkout could not fire. It now asks git which paths are
ignored, and a new check requires bios/ to hold no untracked file.
Both are written as <platform>.json. --manifest-targets aimed at install/
replaced install/retroarch.json with a target table, and --manifest aimed
at install/targets/ did the reverse, both with exit 0. Each mode now stops
when the output directory holds the other kind and names the directory to
use.
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).
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.
unread_on names the platforms whose launcher points the emulator at other files. RetroBat writes absolute BIOS paths into standalone BlastEm's config, so the default names were reported as required and missing there. Such an entry leaves that platform's gap report and pack, and its emulator page says where it is never opened.
A file that arrives with the emulator is already on the user's machine
and the packs leave it out. 4558 collected copies (1.4 GB) were reached
only by such entries, with no platform, other profile entry or install
manifest needing them. They are removed, and two release-asset copies
leave the ignore list. bundled_only.py now lists every copy of a listed
file, since the database keeps one path per content, and pipeline step
4d fails while one remains.
A copy filed under two consoles of one maker was listed twice with the same id, and a data directory listed the same bytes twice, so the site deploy refused 20 pages. The first card keeps the id. The pipeline only generated the pages, so it now renders them with mkdocs build --strict and runs validate_site as the deploy does. The resumed-speed test read the clock and failed on a slow runner, it now fixes the elapsed time.
The platform layer names data trees for its systems: retroarch.yml carries dolphin-sys and ppsspp-assets, and Lakka inherits them, while Lakka's core packages install both trees into the system directory. A profile records that with installed_on on its data directory, and the pack and the install manifest now leave those trees out for that platform.
The data page promises that v1 fields are only ever added, and the envelope schema left items unconstrained, so a renamed field passed validation. Each kind now lists the fields its items carry and their types.
The structured data of an emulator page published one repository and one commit. A profile whose builds live in two repositories, or a port with an upstream_commit, now lists each repository with its own pin.
The collection DAT named its sets "Sony - PlayStation", and the data page said it audits a copy of bios/. clrmamepro and RomVault looked for a folder of that name and reported the 330 sets missing. A Logiqx set name writes a subfolder with a backslash: Sony\PlayStation.
bundled describes the emulator's own release, and a platform package can differ both ways: Batocera's hcl package installs the binary alone and sends users to fetch the data, while its blueMSX core package copies the machines and databases beside the core. not_bundled_on and installed_on name those platforms, and the builder reads them through one rule.
A file installed with its emulator is already on the user's machine and the packs leave it out. bundled_only.py lists the collected copies nothing else needs: no platform declares them, no entry of any build needs them, no install manifest ships them.
A file or tree installed with the emulator is already on the user's machine, so packs, install manifests and verify's pack count leave it out. bundled is scoped by mode like any entry: Dolphin's Sys and PPSSPP's assets ship with the standalone releases, not with the libretro cores, so they now follow the libretro build only and still reach RetroArch packs. PrimeHack, standalone only, ships its own Sys.
The release built and checked the RetroPie pack with two commands after the pipeline, which left it out of the verify and pack consistency check. pipeline.py --include-archived builds and checks it with the rest.
PR validation runs validate_schemas.py --source-only and the pipeline did not, so a profile with an unquoted numeric core name passed every step and failed CI on push. It is step 0 now, skipped where jsonschema is not installed.
Generating the model for every platform met the same two GameIndex.yaml declarations once per platform, and check_freshness opened on six identical warnings.
check_buildbot does not fail the run, yet the summary printed it FAILED right above Pipeline COMPLETE. It now prints WARN and a line saying the difference does not fail the run. The total is aligned under the step durations.
Scrapers resolve their tag or branch head in the constructor, which ran outside the try in scraper_cli: a refused GitHub call ended in a traceback. The error is now printed with the next step, setting GITHUB_TOKEN when the API refused an anonymous call.
The guides said files over 50 MB become release assets and that RPG Maker and ScummVM were the trees kept out of dedup. .gitignore decides the first, four committed files are over 50 MB, and NODEDUP_DIRS holds eleven trees.
With --all, an unknown --target skipped every platform that has targets and ran the others (BizHawk, MiSTer) unfiltered, so verify, slots and the pipeline exited 0 on a typo. A platform without targets is now left out like the rest, and a target no platform has is refused.
The pipeline rebuilds the database with --force, which discarded the previous digests and wrote whatever the CPU computed, on a host that has given an incorrect SHA-1 under load. The forced run still rehashes every file, but a digest that contradicts the cache for an unchanged mtime and size is computed again, and the two readings that agree win, with a warning naming the file.
Two test docstrings named the maintainer's private notes file, and nothing stopped it. They now state the rule itself, and test_prose_style fails on any tracked file that names the private notes or the assistant tooling.
Abdessamad Derraz