Reading this project's own reports gets its own page. Checker code is linked at the revision each platform list was read from. Installer moves to Your setup, legal to Start here, build internals to Contribute, and Reference pages carry a group.
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.
Platform guides, device groups, terminal tabs, pack README manuals and
logo credits move from Python tables to platforms/_guides.yml, checked
against the registry on load. A folder row without a path is the
registry's extract_to, written once.
Wiki pages place themselves in the navigation with order and nav_title
in their front matter. Generated docs/ and mkdocs.yml are unchanged.
Each reference and contributor page now says what it answers and gives
a real case before its rules: scph5501.bin for the check modes and the
data model, DuckStation's entry for profiles, RomM for a platform. The
verification example is the RetroBat d64_1.rom case the gaps page lists
instead of a scph5501.bin case the data cannot produce.
Emulator pages open on what the emulator reads, from the profile, and
keep their technical notes folded. System pages say what a card holds.
The data model shows a real database entry and links the one
resolution chain. Gaps tables name what they count ("Discrepancy",
"Where the files come from"), and the core inventory is titled Cores
by platform so cross-reference keeps one meaning.
validate_site.py fails a page over 1 MB of HTML, a search index over
8.5 MB (1.7 MB gzip), a skipped heading level, a table header without
text, a script from another host and a diagram drawn at runtime. The
limits sit above what the site now measures (0.88 MB, 7.2 MB, 1.43 MB),
each with its reason, and each has a test.
Five tabs from Start here to Contribute, a level on every page, a device picker on Get your files that lists every registered platform, and platform pages that open on where the files go, the command, the pack and how to check. FAQ and troubleshooting keep the questions asked before installing, the technical ones move to their own page, and getting started points to the pages that took its content.
scripts/sitediagrams.py writes the explainer, install flow, verification modes, MinUI card, pipeline and data layer diagrams as inline SVG from the registry, the platform files, the install manifests and the pipeline step list. A page asks for one with an image comment, which the decoration pass resolves. The two mermaid diagrams of the architecture page are replaced, so no page loads mermaid from unpkg.
Platform logos and system icons are copied once by fetch_site_images.py with their source and owner recorded, and the build rewrites the remote links to the copies. validate_site refuses an image from another host. Three registry logos named a personal avatar instead of the project's (RomM, ROCKNIX, MiSTer FPGA) and MyMinUI's was a GitHub identicon, now dropped.
One 1200x630 sharing card per section, a 404 page with search and the two entry pages, and a list of screenshots that need a real device.
No platform YAML holds an md5 shorter than 32 characters: the Batocera scraper completes one when a single collection md5 starts with it. The pages said some entries carry 29 characters and that scrapers store them as-is.
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.
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.
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.
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.
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.
No gradient banner, no coloured stripes or dots beside cards, no lifting cards, pill badges or small-caps headers. File blocks are separated by a rule, tables keep the theme's own size, and headings no longer repeat the site name the header bar already shows. Classes nothing emits were removed.
The documented command, discover tests, makes tests/ the top level,
so tests/__init__.py never ran and the variable it sets to keep tests
offline was missing from every local run. Every documented and CI
run now passes -s tests -t ., and a test requires it.
Both run on Linux but inherited RetroArch's case folding, so a core
file differing from a platform name only by case left their packs:
MesenCE's bios.col and np2kai's lowercase 2608_*.wav, which those
cores open with that exact case. They are now case-sensitive, the
shared groups fold only where the platform does, and Lakka gets its
own pack instead of sharing RetroArch's.
SHA256SUMS.txt sat beside the artifacts it vouches for, so whoever could
rewrite a release rewrote the list with it. The packs were already
reproducible, which answers corruption and lets a third party rebuild an
archive byte for byte; nothing answered a rewritten release.
The list is now signed with an ed25519 key kept for this alone, and the
public half is allowed_signers at the repository root, so verification
does not go through the release page: ssh-keygen -Y verify against the
committed file, then sha256sum --check. Rehearsed on all three outcomes:
a good signature, a tampered pack caught by the sums, a rewritten list
caught by the signature.
The release steps sign and upload the signature, the README points a
downloader at the procedure, and the reproducibility section says what
each half proves. Rotation keeps retired lines so past releases stay
verifiable. Three tests hold the trust root, the signing step and the
documented principal in agreement.
validate.yml triggered on pull_request alone, and it holds the only
unittest invocation in the repository: deploy-site.yml stops at
validate_schemas, generation and the freshness diff. Work lands on main
by direct push far more often than by pull request, so 1,318 cases were
guarding the road almost nothing takes.
The suite and the schema check now run on both events. validate-bios and
label-pr read pull request context and carry an event guard. The
concurrency group falls back to the ref, so a push series collapses to
the tip: what stays verified is the head of main.
The path lists are spelled out per event because the workflow parser
reads no YAML anchor, which PyYAML would have accepted in silence. Four
tests hold the wiring: the suite reachable from a push, the two path
lists equal, every job reading pull request context guarded, and no
anchor in any workflow.
Pinning every member's metadata made packs reproducible and took the
executable bit with it. The RetroDECK pack ships the two Voxatron engine
binaries, and extracted at 644 they cannot be run.
Git records the bit, so reading it from the source file keeps a pack the
same from any clone. Nothing else about the source's mode reaches the
archive: 2569 members ship at 644 and 942 at 755, which is what the
builder produced before the pinning.
Nothing caught this. The comparison that proved the pinning inert
checked member names, CRCs and sizes, and mode is none of those. A test
now builds a runnable payload and asserts it survives extraction.
RetroDECK rebuilds to the same bytes twice and passes its integrity
check, 2008/2008 baseline and 1551/1551 cores.
The job carried if: false, which blocks workflow_dispatch as well as
push, so there was no way to cut a release through CI at all and the
comment described a temporary state that had become permanent.
Releasing is deliberate: the push trigger is gone and the job runs when
someone dispatches it. The seven-day rate limit stays as the guard
against dispatching twice, and concurrency no longer cancels a run that
may be midway through uploading assets.
The pages described workflows, an installer pinning and a CI permission
model that are not the ones in the tree. Corrected: the release process,
the CI table, the resolution chain and its statuses, the hash-mismatch
policy per verification mode, the archive convention, and the romset
recipe store.
The FAQ keeps the project's own reading of the legal question rather
than a version that reaches no conclusion.
Add and reorder BIOS path entries in the site generator (BizHawk, EmuDeck, RetroPie, RomM). Update the add-platform pipeline steps and CI workflow notes. Document verification behavior changes: FirmwareDatabase index now includes sha256; RomM uses MD5 verification (verify.py checks MD5 only); BizHawk uses SHA1; severity label for GREEN adjusted to WARNING. Clarify troubleshooting/verify output semantics (UNTESTED and mismatch reporting), add profiling fields (core_classification option and adler32), fix several path and link typos (RetroDECK path, README/CONTRIBUTING links), and other small docs polishing.
Update wiki source files (the single source of truth for the site):
- tools.md: renumber pipeline steps 1-8, add step 6 (pack integrity),
add missing CLI flags for cross_reference.py and refresh_data_dirs.py
- architecture.md: update mermaid diagram with pack integrity step,
fix test file count (5 files, 249 tests)
- testing-guide.md: add test_pack_integrity section, add step 5 to
verification discipline checklist
9 new wiki pages: getting-started, faq, troubleshooting,
advanced-usage, verification-modes, adding-a-platform,
adding-a-scraper, testing-guide, release-process.
Updated architecture.md with mermaid diagrams, tools.md with
full pipeline and target/exporter sections, profiling.md with
missing fields, index.md with glossary and nav links.
Expanded CONTRIBUTING.md from stub to full contributor guide.
Filter .old.yml from load_emulator_profiles, generate_db alias
collection, and generate_readme counts. Fix BizHawk sha1 mode
in tools.md, fix RetroPie path, fix export_truth.py typos.
Abdessamad Derraz