chore: merge origin/main

.github/workflows/deploy-site.yml

@@ -17,6 +17,7 @@ on:
       - "scripts/cross_reference.py"
       - "scripts/crypto_verify.py"
       - "scripts/derivations.py"
+      - "scripts/dedup.py"
       - "scripts/deterministic_zip.py"
       - "scripts/dumpcatalog.py"
       - "scripts/exporter/baseline.py"
@@ -61,6 +62,7 @@ on:
       - "scripts/split_pack.py"
       - "scripts/truth.py"
       - "scripts/upstream.py"
+      - "scripts/validate_pr.py"
       - "scripts/validate_schemas.py"
       - "scripts/validate_site.py"
       - "scripts/validation.py"

CONTRIBUTING.md

@@ -24,8 +24,12 @@ the most useful contribution.
 
 From a clone, the file goes to `bios/Manufacturer/Console/filename`, and
 `python scripts/validate_pr.py path/to/file` names that folder for a file still
-outside `bios/`. Given a folder, such as an old card's BIOS folder, it checks
+outside `bios/`, and `--place` copies it there under its declared name. Given a
-every file under it. A file over the browser's 25 MB limit goes in that way.
+folder, such as an old card's BIOS folder, it checks every file under it. A
+file over the browser's 25 MB limit goes in that way, up to the 100 MB GitHub
+accepts in a push. A larger file stays out of the pull request: [open an
+issue](https://github.com/Abdess/retrobios/issues/new?template=missing-bios.yml) with the hashes `validate_pr.py` prints, and a
+maintainer adds it to the `large-files` release.
 
 ## Add a platform
 
@@ -64,18 +68,22 @@ validation above is its check. For code, profiles or platform files, these are
 the checks CI runs on a pull request (Python 3.12 or newer):
 
 ```bash
-pip install pyyaml "jsonschema[format-nongpl]" markdown
+pip install pyyaml "jsonschema[format-nongpl]" markdown pyte hypothesis vermin
 python scripts/validate_pr.py --changed          # the BIOS files the branch adds
 python scripts/validate_schemas.py --source-only # profiles and platform files
 python scripts/profile_sync.py --added --base upstream/main --strict  # each profile the branch adds
 python scripts/refresh_data_dirs.py  # data directories the pack-building tests read
 python scripts/generate_site.py       # the pages the site tests read
+npm ci --prefix website && npm run build --prefix website  # the built site, Node 24
 python -m unittest discover -s tests -t .
+python scripts/validate_site.py --site-dir website/dist  # links and page budgets
+vermin --no-tips -t=3.8- --violations install.py         # install.py stays on Python 3.8
 ```
 
 `upstream` is this repository as a remote of your fork
-(`git remote add upstream https://github.com/Abdess/retrobios`). A profile you
+(`git remote add upstream https://github.com/Abdess/retrobios`, then
-only edit is not held to `--strict`: its older refs are reviewed by hand.
+`git fetch upstream`). A profile you only edit is not held to `--strict`: its
+older refs are reviewed by hand.
 
 The full pipeline is not needed for a contribution: it rebuilds every pack
 in `dist/`, which is why maintainers run it before a release.

emulators/gearboy.yml

@@ -19,8 +19,9 @@ systems:
 
 notes: |
   Gearboy is a Game Boy / Game Boy Color emulator by drhelius (Ignacio Sanchez).
-  The buildbot builds from drhelius/Gearboy (upstream), not from libretro/Gearboy
+  The buildbot builds from drhelius/Gearboy (upstream), not from libretro/Gearboy,
-  (abandoned fork frozen at v2.6.0 from 2018).
+  an archived fork (its description reads "NOTE: PUBLIC ARCHIVE") whose
+  src/definitions.h states version 2.6.0.
   Bootrom loading is gated behind core options (disabled by default).
   Without bootroms, the core HLE-skips the boot animation.
 

emulators/np2kai.yml

@@ -16,7 +16,7 @@ cores: [np2kai]
 systems: [pc-98]
 
 # NP2kai is a PC-9801/9821 emulator by AZO234, enhanced fork of Neko Project II
-# by Yui (original upstream dead: np2.yui.ne.jp). All files load from
+# by Yui (original site: np2.yui.ne.jp). All files load from
 # <system>/np2kai/ subdirectory, set in retro_load_game()
 # (sdl/libretro/libretro.c:1800-1815). All getbiospath() calls resolve
 # relative to that directory.

emulators/pcsx1.yml

@@ -15,7 +15,7 @@ display_name: "Sony - PlayStation (PCSX1)"
 cores: [pcsx1]
 systems: [sony-playstation]
 notes: >
-  Never-completed fork of PCSX-ReARMed targeting general PC hardware. Experimental, abandoned since 2018.
+  Fork of PCSX-ReARMed targeting general PC hardware. The last commit on libretro/pcsx1-libretro is from 2018-02-11.
   HLE BIOS built-in: if no BIOS file found, core warns and falls back to HLE automatically.
   BIOS search order: scph1001.bin, scph5501.bin, scph7001.bin by exact filename, then scans system
   directory for any file matching "scph*" (case-insensitive) with size == 512 KB.

emulators/rustation.yml

@@ -30,7 +30,7 @@ notes: >
   Database contains 24 entries (22 good, 2 known-bad rejected)
   covering BIOS versions 1.0 through 4.5.
   .info declares firmware_count=0, while the core reads the BIOS images this profile lists.
-  Project abandoned, superseded by rustation-ng.
+  Its README says it is no longer developed and points to rustation-ng.
 
 files:
   # -- Region: Japan --

emulators/yabause.yml

@@ -17,9 +17,9 @@ systems:
   - sega-saturn
 
 notes: |
-  Yabause is a software-rendered Sega Saturn emulator. Upstream project is
+  Yabause is a software-rendered Sega Saturn emulator. The last commit on the
-  dead. The libretro port is maintained by the community on the master branch of
+  master branch of Yabause/yabause is from 2018-07-07. The libretro port is
-  libretro/yabause.
+  maintained by the community on the master branch of libretro/yabause.
 
   Saturn BIOS search order:
     1. {system}/saturn_bios.bin

scripts/cross_reference.py

@@ -27,6 +27,7 @@ from common import (
     platform_arg,
     ProfileSelectionError,
     _norm_system_id,
+    archive_entry,
     directory_members,
     expand_directory_entries,
     get_mame_clone_map,
@@ -573,18 +574,27 @@ class FileConsumers:
                         for dest, sha1 in members.items()
                     ]
                 else:
-                    # The archive members the profile reads decide between
+                    found = [self.sha1_of(*entry_target(entry, files), owner=name)]
-                    # same-named copies, as in the emulator pack and verify.
-                    found = [self.sha1_of(
-                        with_archive_members(entry, files), entry.get("path") or "",
-                        owner=name,
-                    )]
                 for sha1 in found:
                     if sha1:
                         self.emulators.setdefault(sha1, set()).add(name)
                         self.emulator_entries.setdefault(sha1, {}).setdefault(name, index)
 
 
+def entry_target(entry: dict, files: list[dict]) -> tuple[dict, str]:
+    """The entry a profile entry resolves through, and its destination.
+
+    A ROM read inside an archive is the archive: FBNeo opens sm1.sm1 in
+    neogeo.zip, and resolved by its crc32 it credited a loose copy FBNeo
+    never opens, while no neogeo.zip card named FBNeo. The archive
+    members the profile reads decide between same-named copies, as in
+    the emulator pack and verify.
+    """
+    if entry.get("archive"):
+        return archive_entry(entry["archive"], files), ""
+    return with_archive_members(entry, files), entry.get("path") or ""
+
+
 def entry_source(f: dict, index: dict) -> str | None:
     """Where the collection holds a profile entry, or None.
 

scripts/generate_readme.py

@@ -37,6 +37,7 @@ from common import (
 from nativemode import check_label
 import release_record
 from siteguides import install_command
+from validate_pr import LARGE_FILE_ISSUE_URL
 from verify import verify_platform
 
 
@@ -929,8 +930,12 @@ the most useful contribution.
 
 From a clone, the file goes to `bios/Manufacturer/Console/filename`, and
 `python scripts/validate_pr.py path/to/file` names that folder for a file still
-outside `bios/`. Given a folder, such as an old card's BIOS folder, it checks
+outside `bios/`, and `--place` copies it there under its declared name. Given a
-every file under it. A file over the browser's 25 MB limit goes in that way.
+folder, such as an old card's BIOS folder, it checks every file under it. A
+file over the browser's 25 MB limit goes in that way, up to the 100 MB GitHub
+accepts in a push. A larger file stays out of the pull request: [open an
+issue]({LARGE_FILE_ISSUE_URL}) with the hashes `validate_pr.py` prints, and a
+maintainer adds it to the `large-files` release.
 
 ## Add a platform
 
@@ -969,18 +974,22 @@ validation above is its check. For code, profiles or platform files, these are
 the checks CI runs on a pull request (Python 3.12 or newer):
 
 ```bash
-pip install pyyaml "jsonschema[format-nongpl]" markdown
+pip install pyyaml "jsonschema[format-nongpl]" markdown pyte hypothesis vermin
 python scripts/validate_pr.py --changed          # the BIOS files the branch adds
 python scripts/validate_schemas.py --source-only # profiles and platform files
 python scripts/profile_sync.py --added --base upstream/main --strict  # each profile the branch adds
 python scripts/refresh_data_dirs.py  # data directories the pack-building tests read
 python scripts/generate_site.py       # the pages the site tests read
+npm ci --prefix website && npm run build --prefix website  # the built site, Node 24
 python -m unittest discover -s tests -t .
+python scripts/validate_site.py --site-dir website/dist  # links and page budgets
+vermin --no-tips -t=3.8- --violations install.py         # install.py stays on Python 3.8
 ```
 
 `upstream` is this repository as a remote of your fork
-(`git remote add upstream https://github.com/Abdess/retrobios`). A profile you
+(`git remote add upstream https://github.com/Abdess/retrobios`, then
-only edit is not held to `--strict`: its older refs are reviewed by hand.
+`git fetch upstream`). A profile you only edit is not held to `--strict`: its
+older refs are reviewed by hand.
 
 The full pipeline is not needed for a contribution: it rebuilds every pack
 in `dist/`, which is why maintainers run it before a release.

scripts/generate_site.py

@@ -63,7 +63,7 @@ from common import (
 )
 from artifacts import ArtifactLockBusy, artifact_lock
 from artifacts import json_text as _json_text
-from cross_reference import FileConsumers, held_copy
+from cross_reference import FileConsumers, entry_target, held_copy
 from nativemode import check_label, digest_algorithm, reads_file_contents
 
 
@@ -679,9 +679,13 @@ def release_line(platform_name: str, record: dict) -> list[str]:
     return lines
 
 
-# What a reader on another version of the setup does with the pack. Our data
+def other_version(platform_name: str, prefix: str = "../") -> str:
-# says which list the pack was made from, not what an older list lacks.
+    """What a reader on another version of the setup does with the pack.
-OTHER_VERSION = "Other version? Run `--check` after copying."
+    Our data says which list the pack was made from, not what an older list
+    lacks. A bare `--check` was no command to run for someone holding a ZIP:
+    the line links the setup page's commands that check."""
+    target = f"{prefix}platforms/{platform_name}.md#how-to-check"
+    return f"Other version? [Check the files]({target}) after copying."
 
 
 def setup_version(platform_name: str, record: dict | None = None) -> str:
@@ -700,13 +704,15 @@ def setup_version(platform_name: str, record: dict | None = None) -> str:
     return f"{config.get('platform', platform_name)} {version.removeprefix('v')}"
 
 
-def version_line(platform_name: str, record: dict, named: bool = False) -> str:
+def version_line(platform_name: str, record: dict, named: bool = False,
+                 prefix: str = "../") -> str:
     """The version the pack fits, and what to do on another. *named* when
     the release line already named the pack's version."""
     made_from = setup_version(platform_name, record)
+    other = other_version(platform_name, prefix)
     if named or not made_from:
-        return OTHER_VERSION
+        return other
-    return f"Made for {made_from}. {OTHER_VERSION}"
+    return f"Made for {made_from}. {other}"
 
 
 # Why a setup has no ZIP, answered once on the FAQ.
@@ -793,7 +799,8 @@ def _pack_buttons(platform_name: str, label: str = "Pack", prefix: str = "../",
     )
     released = release_line(platform_name, record)
     return "\n\n".join([buttons, *released,
-                        version_line(platform_name, record, named=len(released) > 1)])
+                        version_line(platform_name, record, named=len(released) > 1,
+                                     prefix=prefix)])
 
 
 def generate_get_page(platform_name: str) -> str:
@@ -2191,9 +2198,11 @@ def generate_data_page(exports: list[dict], measurements_sqlite: Path | None = N
         "when one is known, the `system_name` the site gives its system. "
         "[Find or check a file](files/index.md) reads it.",
         "",
-        "In a browser, the SHA1 of a file the reader picks finds its records. Web "
+        "In a browser, the SHA1 of a file the reader picks finds its records when "
-        "Crypto has no MD5, and none is needed: a setup that lists a file by MD5 "
+        "the collection holds that file. A list naming a file the collection lacks "
-        "or CRC32 is filed under the SHA1 of the collection file it names too.",
+        "is filed under its MD5 or CRC32 alone, and Web Crypto has no MD5: "
+        "that page computes MD5 and CRC32 in script and looks "
+        "up all four.",
         "",
         "```js",
         "// The records of the file a reader picked, by its SHA1.",
@@ -2443,7 +2452,7 @@ def _browser_title(relative: Path, title: str) -> str:
     # The words a web search for a console or an emulator types: "dreamcast
     # bios", "duckstation bios". The h1 and the sidebar keep the name.
     if key.startswith("emulators/"):
-        has_files = PAGE_DATA.get(key, {}).get("hasPart")
+        has_files = PAGE_DATA.get(key, {}).get("hasPart") and key not in NOTHING_TO_FETCH
         # Lexaloffle, QEMU and ScummVM name a system page too.
         own = " emulator" if f"systems/{relative.name}" in PAGE_DATA else ""
         return f"{title}{own} BIOS files" if has_files else f"{title} emulator"
@@ -2514,6 +2523,12 @@ def _page_description(markdown: str, relative: Path, title: str) -> str:
         return _system_description(markdown, browser)
     if key.startswith("emulators/") and relative.stem != "index":
         return _emulator_description(markdown, key, title, _first_paragraph(lead))
+    if key.startswith("platforms/") and relative.stem != "index":
+        # The summary ends on a link to the platform's site, whose label
+        # alone was left: "A system for handhelds. ROCKNIX website."
+        summary = _first_paragraph(lead).removesuffix(f" {title} website.")
+        return _shorten(f"The BIOS files for {title}, the folder they go in and the "
+                        f"command that installs them. {summary}", 155)
     first = _first_paragraph(lead, minimum=35)
     questions = [_plain_markdown(re.sub(r"\s*\{[^}]*\}\s*$", "", heading))
                  for heading in re.findall(r"^## (.+\?)(?:[ \t]*\{[^}]*\})?[ \t]*$", markdown, re.M)]
@@ -2539,7 +2554,9 @@ def _emulator_description(markdown: str, key: str, title: str, intro: str) -> st
     its profile table when the profile names no system."""
     notes = re.search(r"<summary>Technical notes</summary>\s*\n\n(.+?)(?:\n\n|$)", markdown, re.S)
     if not intro and notes:
-        sentence = re.split(r"(?<=\.)\s", _plain_markdown(notes[1]), maxsplit=1)[0]
+        # An initial is no sentence end: hcl's read "open-source port of E."
+        sentence = re.split(r"(?<![\s(][A-Z]\.)(?<=\.)\s", _plain_markdown(notes[1]),
+                            maxsplit=1)[0]
         intro = f"{title}: {sentence[:1].lower()}{sentence[1:]}"
     if not intro:
         row = {name: _plain_markdown(value) for name, value in re.findall(
@@ -2553,6 +2570,8 @@ def _emulator_description(markdown: str, key: str, title: str, intro: str) -> st
     elif title not in intro:
         # "It reads 1 file": the profile names no system to open on.
         intro = f"{title}: {intro[:1].lower()}{intro[1:]}"
+    if key in NOTHING_TO_FETCH:
+        return _shorten(f"{intro} Every file it reads comes with the app.", 155)
     # An alias page sends the reader to the core whose files it shares.
     if PAGE_DATA.get(key, {}).get("hasPart") or "uses the same firmware as" in intro:
         return _shorten(intro, 155)
@@ -2623,6 +2642,9 @@ def _system_description(markdown: str, browser: str) -> str:
 # Structured data a page generator knows beyond its text, by page path
 # relative to docs/: read by decorate_markdown_pages.
 PAGE_DATA: dict[str, dict] = {}
+# Emulator pages whose every file comes with the app: no "BIOS files" in
+# their title, and their description says so (DaedalusX64, ZEsarUX).
+NOTHING_TO_FETCH: set[str] = set()
 # Pages that hold part of a page too long for one (a console of a split
 # manufacturer, a run of an emulator's cards), by path relative to docs/:
 # their H1 already says what they are.
@@ -2950,6 +2972,17 @@ def release_versions(profile: dict) -> dict[str, str]:
             if re.search(r"[0-9]", value) and not names_a_commit(value)}
 
 
+def record_emulator_page(name: str, profile: dict) -> None:
+    """What the title and the description of an emulator page read."""
+    key = f"emulators/{name}.md"
+    PAGE_DATA[key] = emulator_structured_data(profile)
+    listed = [f for f in profile.get("files") or [] if isinstance(f, dict)]
+    if listed and all(comes_with_app(f) for f in listed):
+        NOTHING_TO_FETCH.add(key)
+    else:
+        NOTHING_TO_FETCH.discard(key)
+
+
 def emulator_structured_data(profile: dict) -> dict:
     """What an emulator page states, for a reader of its JSON-LD.
 
@@ -4176,7 +4209,7 @@ def _setup_routes(name: str, display: str, page_link: bool = False) -> list[str]
             # lines are on the get/ page, the one the command is shared from.
             tabs = ("windows",) if page_link else ("windows", "macos", "linux")
             computer += [
-                "Needs the device on your network. Run the command on the computer.",
+                "Needs the device on your network.",
                 "",
                 *terminal_tabs(tabs, args, how=how),
                 *([f"[Mac or Linux](../{get_page_link(name)}#from-your-computer)", ""]
@@ -6443,13 +6476,6 @@ def _region_order() -> dict[str, int]:
     return {region: index for index, region in enumerate(ordered)}
 
 
-_REGION_WORDS = {"uk": "UK", "asia-ntsc": "Asia (NTSC)", "asia-pal": "Asia (PAL)"}
-
-
-def _region_word(region: str) -> str:
-    return _REGION_WORDS.get(region) or region.replace("-", " ").title()
-
-
 def _reading_pairs(sha1: str, consumers: FileConsumers) -> list[tuple[str, dict]]:
     """The profile entries that read a file, each with its emulator."""
     files_of = getattr(consumers, "emulator_files", {})
@@ -6662,7 +6688,7 @@ def _start_groups(
     for f in primaries:
         sha1 = str(f.get("sha1", ""))
         for emulator, entry in _reading_pairs(sha1, consumers):
-            if entry.get("bundled"):
+            if comes_with_app(entry):
                 continue
             system = None
             systems = _entry_systems(emulator, entry)
@@ -6677,7 +6703,7 @@ def _start_groups(
         for sha1, emulator, entry, system in _entries_by_section(consumers).get(here, []):
             record = known.get(sha1)
             parts = str((record or {}).get("path", "")).split("/")
-            if (sha1 in held or not record or entry.get("bundled") or ".variants" in parts
+            if (sha1 in held or not record or comes_with_app(entry) or ".variants" in parts
                     or len(parts) < 3 or parts[1] != "Arcade"):
                 continue
             if sha1 not in kept:
@@ -7031,7 +7057,7 @@ def _group_start(group: _StartGroup, first_below: bool) -> list[tuple[str, str]]
         else:
             regional[(*same, region)] = regional.pop(same)
 
-    out = [(", ".join(_region_word(r) for r in regions), regional[regions])
+    out = [(", ".join(_region_words(r) for r in regions), regional[regions])
            for regions in sorted(regional, key=lambda rs: (order.get(rs[0], len(order)), rs))]
     # A name a region line gives already is no "also", and one only
     # emulators no region line answers reads is an "instead": armsx1 reads
@@ -7819,12 +7845,15 @@ def generate_emulators_index(profiles: dict) -> str:
 
 
 def comes_with_app(f: dict) -> bool:
-    """The emulator installs the file in every build the profile documents.
+    """The emulator installs the file in the build the entry describes.
 
-    ``bundled`` narrowed by ``mode`` is installed by one build only, and the
+    The rule verify and the pack read (``installed_with_emulator``). An entry
-    other build still reads it from the user's folder.
+    narrowed by ``mode`` is that build's alone: the other build reads its
+    own entry, or nothing.
     """
-    return f.get("bundled") is True and not f.get("mode")
+    from validation import installed_with_emulator  # noqa: PLC0415
+
+    return installed_with_emulator(f)
 
 
 def _file_badges(f: dict, in_repo: bool, for_system: str = "") -> list[str]:
@@ -7841,7 +7870,6 @@ def _file_badges(f: dict, in_repo: bool, for_system: str = "") -> list[str]:
     category = f.get("category", "")
     region = f.get("region", "")
     storage = f.get("storage", "")
-    bundled = f.get("bundled", False)
     embedded = f.get("embedded", False)
     has_builtin = f.get("has_builtin", False)
     archive = f.get("archive", "")
@@ -7851,9 +7879,10 @@ def _file_badges(f: dict, in_repo: bool, for_system: str = "") -> list[str]:
     if shipped:
         # NetherSX2's three interface fonts sat beside the BIOS as
         # "required" and "missing": the app ships them, nobody fetches them.
-        badges.append(
+        # Narrowed to one build, the build is named.
-            '<span class="rb-badge rb-badge-info">comes with the app</span>'
+        build = f.get("mode") if f.get("mode") in ("standalone", "libretro") else ""
-        )
+        label = f"installed with the {build} emulator" if build else "comes with the app"
+        badges.append(f'<span class="rb-badge rb-badge-info">{label}</span>')
     elif required:
         needed = f"required for {html.escape(for_system)}" if for_system else "required"
         badges.append(
@@ -7879,7 +7908,7 @@ def _file_badges(f: dict, in_repo: bool, for_system: str = "") -> list[str]:
         badges.append(
             '<span class="rb-badge rb-badge-info">HLE fallback</span>'
         )
-    if mode:
+    if mode and not (shipped and mode in ("standalone", "libretro")):
         badges.append(
             f'<span class="rb-badge rb-badge-muted">{mode}</span>'
         )
@@ -7896,13 +7925,6 @@ def _file_badges(f: dict, in_repo: bool, for_system: str = "") -> list[str]:
         badges.append(
             f'<span class="rb-badge rb-badge-muted">{storage}</span>'
         )
-    if bundled and not shipped:
-        # Said in words: the pack leaves the file out because the emulator,
-        # in the build the entry names, already installs it.
-        build = f" {f['mode']}" if f.get("mode") in ("standalone", "libretro") else ""
-        badges.append(
-            f'<span class="rb-badge rb-badge-muted">installed with the{build} emulator</span>'
-        )
     if embedded:
         badges.append(
             '<span class="rb-badge rb-badge-muted">embedded</span>'
@@ -7995,10 +8017,14 @@ def _build_rows(profile: dict) -> list[str]:
     found = _found_by_text(profile.get("found_by"))
     if found:
         rows.append(f"| Files found by | {found} |")
-    frozen = profile.get("pin_frozen")
+    # Each pin says why it is held: a bare upstream revision read as a
-    if frozen:
+    # profile left behind HEAD (mupen64plus, aps3e).
-        held = _escape_tags(' '.join(str(frozen).split())).replace('|', r'\|')
+    for field, label in (("pin_frozen", "Revision held"),
-        rows.append(f"| Revision held | {held} |")
+                         ("upstream_frozen", "Upstream revision held")):
+        frozen = profile.get(field)
+        if frozen:
+            held = _escape_tags(' '.join(str(frozen).split())).replace('|', r'\|')
+            rows.append(f"| {label} | {held} |")
     announced = profile.get("version_source")
     if isinstance(announced, dict) and announced.get("url"):
         url = str(announced["url"])
@@ -8193,10 +8219,13 @@ def _render_emulator_file(
     if consumers and f.get("type") == "directory":
         details.extend(_directory_copies(f, consumers))
     if consumers:
-        sha1 = consumers.sha1_of(f, f.get("path") or "", owner=owner)
+        sha1 = consumers.sha1_of(
+            *entry_target(f, profile.get("files") or []), owner=owner)
         held = consumers.path_of(sha1) if sha1 and in_repo else ""
         parts = held.split("/")
-        if len(parts) >= 3 and parts[0] == "bios":
+        if len(parts) >= 3 and parts[0] == "bios" and f.get("archive"):
+            details.append(_held_copy_line((held, f["name"]), consumers))
+        elif len(parts) >= 3 and parts[0] == "bios":
             # Which copy answers this entry, linked to its collection card
             # (provenance, every hash, download): an entry declaring no hash
             # said "in repo" without naming the file.
@@ -8473,8 +8502,9 @@ def emulator_romset(profile: dict) -> str:
     return ""
 
 
-def _version_said(value: object, basis: dict) -> str:
+def _version_said(value: object, basis: dict, profile: dict, mode: str = "") -> str:
-    """One build's version and where the profile reads it."""
+    """One build's version and where the profile reads it, linked to the
+    file or the tag where the source allows."""
     kind = basis.get("kind")
     if kind == "none":
         return "none stated in the source"
@@ -8483,13 +8513,18 @@ def _version_said(value: object, basis: dict) -> str:
             return f"unknown: {_UNVERIFIED.get(str(basis.get('why')), 'unread')}"
         return "none published"
     value = str(value)
+    url = version_basis_url(profile, basis, mode) if kind in ("file", "tag") else ""
+
+    def shown(text: object) -> str:
+        return f"[{_code(str(text))}]({url})" if url else _code(str(text))
+
     if kind == "file":
-        return f"{value}, stated in `{basis.get('path')}`"
+        return f"{value}, stated in {shown(basis.get('path'))}"
     if kind == "tag":
         distance = int(basis.get("distance") or 0)
         if not distance:
             return f"{value}, the tag at this revision"
-        return f"{value}, {distance} commit{'s' if distance != 1 else ''} after tag `{basis.get('tag')}`"
+        return f"{value}, {distance} commit{'s' if distance != 1 else ''} after tag {shown(basis.get('tag'))}"
     if kind == "package":
         return f"{value}, read from {basis.get('artifact')}"
     if kind == "unverified":
@@ -8524,10 +8559,10 @@ def version_markdown(profile: dict) -> str:
         values = value if isinstance(value, dict) else {}
         modes = list(dict.fromkeys([*values, *basis]))
         return ", ".join(
-            f"{_version_said(values.get(mode), basis.get(mode) or {})} ({mode})"
+            f"{_version_said(values.get(mode), basis.get(mode) or {}, profile, mode)} ({mode})"
             for mode in modes
         )
-    return _version_said(value, basis)
+    return _version_said(value, basis, profile)
 
 
 ROMSET_FAQ_PAGE = Path(WIKI_SRC_DIR) / "technical-faq.md"
@@ -8860,7 +8895,7 @@ def standalone_place(name: str, profile: dict, entry: dict, key: str = "") -> st
     if not isinstance(entry, dict) or not profile.get("data_root"):
         return ""
     mode = entry.get("mode")
-    if mode == "libretro" or (entry.get("bundled") is True and mode in (None, "standalone")):
+    if mode == "libretro" or comes_with_app(entry):
         return ""
     excluded = _folder_only(name, entry)
     if "*" in excluded or (key in excluded if key else excluded):
@@ -8898,7 +8933,7 @@ def _data_root_folder(name: str, profile: dict, key: str, entries: list[dict]) -
     folders = set()
     for f in entries:
         mode = f.get("mode")
-        if mode == "libretro" or (f.get("bundled") is True and mode in (None, "standalone")):
+        if mode == "libretro" or comes_with_app(f):
             continue
         place = standalone_place(name, profile, f, key)
         if not place:
@@ -12142,15 +12177,16 @@ def _generate(args: argparse.Namespace) -> None:
     # The test each emulator page builds for itself: the parts a page is
     # cut into weigh the status its rows show.
     file_held = _availability_check(db, suppl_names)
+    # The plan names systems: the sections must be this run's data first.
+    manufacturers = _group_by_manufacturer(db)
+    set_console_sections(manufacturers, program_names(profiles))
     plan_emulator_pages(profiles, file_held)
     # Build system_id -> manufacturer page map (needed by all generators)
     print("Building system cross-reference map...")
-    manufacturers = _group_by_manufacturer(db)
     set_system_page_map(
         _system_pages_from_data(manufacturers, coverages, db, profiles)
     )
     print(f"  {len(_system_page_map)} system IDs mapped to pages")
-    set_console_sections(manufacturers, program_names(profiles))
     from siterender import link_coverage, system_ids  # noqa: PLC0415
 
     print("Generating system pages...")
@@ -12300,7 +12336,7 @@ def _generate(args: argparse.Namespace) -> None:
             name, profile, db, consumers, suppl_names, data_registry,
             data_dirs=page_data_directories(name, profiles),
         )
-        PAGE_DATA[f"emulators/{name}.md"] = emulator_structured_data(profile)
+        record_emulator_page(name, profile)
         # A frozen copy answers after the emulator: a frozen_snapshot, and a
         # profile pinned to the revision one platform builds (pin_frozen).
         # Reicast and Flycast 2021 LE answered "dc bios missing" before Flycast.
@@ -12466,6 +12502,7 @@ from sitesources import (  # noqa: E402,F401
     _forge_sources,
     _source_permalink,
     _source_ref_markdown,
+    version_basis_url,
 )
 from siterender import (  # noqa: E402,F401
     _by_mode,

scripts/pipeline.py

@@ -35,8 +35,8 @@ come from .cache/large/ only.
 Usage:
     python scripts/pipeline.py                    # active platforms
     python scripts/pipeline.py --include-archived # all platforms
-    python scripts/pipeline.py --skip-packs       # steps 0 to 3b, 4d, 7 and 8
+    python scripts/pipeline.py --skip-packs       # skip steps 4, 4b, 4c, 5 and 6
-    python scripts/pipeline.py --skip-docs        # skip steps 7 and 8
+    python scripts/pipeline.py --skip-docs        # skip steps 7, 8 and 8b
     python scripts/pipeline.py --offline          # no network, see above
 """
 
@@ -747,13 +747,13 @@ def main() -> None:
     parser.add_argument(
         "--skip-packs",
         action="store_true",
-        help="Skip steps 4 to 6 except 4d: packs, manifests and their checks. "
+        help="Skip steps 4, 4b, 4c, 5 and 6: packs, manifests and their checks. "
-        "Steps 7 and 8 still rewrite README.md and the site pages in website/, "
+        "Step 4d still gates the run, step 4e still rewrites install/games/, and "
-        "add --skip-docs to "
+        "steps 7 to 8b rewrite README.md and the site unless --skip-docs is given",
-        "stop after verify",
     )
     parser.add_argument(
-        "--skip-docs", action="store_true", help="Skip README and site generation"
+        "--skip-docs", action="store_true",
+        help="Skip steps 7, 8 and 8b: README, installer page and site"
     )
     parser.add_argument(
         "--offline", action="store_true",

scripts/profile_sync.py

@@ -4378,6 +4378,10 @@ def branch_profiles(emulators_dir: str, base: str, added_only: bool) -> list[str
     branch point are not counted, and ends at the working tree, so a
     profile not yet committed is checked too. A rename reads as an addition.
     """
+    if subprocess.run(["git", "rev-parse", "--verify", "--quiet", base],
+                      capture_output=True, check=False).returncode:
+        remote = base.split("/", 1)[0] if "/" in base else "upstream"
+        raise SystemExit(f"ERROR: {base} is not in this clone: run git fetch {remote} first")
     merge_base = _git_lines(["merge-base", base, "HEAD"])[0]
     paths = _git_lines([
         "diff", "--name-only", "--no-renames",

scripts/siteguides.py

@@ -454,6 +454,10 @@ def fence(shell: str, command: str, indent: str = "") -> list[str]:
     return [f'{indent}```{shell} title="{FENCE_TITLES[shell]}"', f"{indent}{command}", f"{indent}```"]
 
 
+# Terminals named by where the command runs, not by the system it runs on.
+PLACES = ("device", "server")
+
+
 def terminal_tabs(
     tabs: tuple[str, ...], args: dict[str, str], indent: str = "", steps: bool = True,
     how: dict[str, str] | None = None,
@@ -465,6 +469,8 @@ def terminal_tabs(
     items = []
     for key in tabs:
         label, shared, shell = TERMINALS[key]
+        if len(tabs) == 1 and key in PLACES:
+            label = ""  # one place, which the text around already names
         text = (how or {}).get(key) or shared
         body = [text, ""] if steps else []
         items.append((label, body + fence(shell, install_command(shell, args.get(key, "")))))

scripts/sitemarkup.py

@@ -139,8 +139,17 @@ def details(kind: str, title: str, body: list[str], indent: str = "", opened: bo
 
 def tabs(items: list[tuple[str, list[str]]], indent: str = "", sync: str = "os") -> list[str]:
     """Content tabs. Tabs that share a label share the reader's choice
-    across pages."""
+    across pages.
+
+    One item is no choice: 142 bars offered a single tab, most of them
+    "On the device" under a heading that says so. Its body stands alone,
+    led by its label when it has one: the RetroPie card is read from Linux
+    only, and the label is where the page said so."""
     lines: list[str] = []
+    if len(items) == 1:
+        label, body = items[0]
+        lead = [f"{indent}**{label}**", ""] if label else []
+        return [*lead, *_indented(_trim(body), indent), ""]
     if _engine == "mkdocs":
         for label, body in items:
             lines += [f'{indent}=== "{label}"', "", *_indented(_trim(body), indent + "    "), ""]

scripts/sitesources.py

@@ -240,8 +240,15 @@ def _walk(value: object) -> Iterator[tuple[str, object]]:
 
 def cited_paths(profile: dict) -> set[str]:
     """Every path a page of the profile could link: the parts of each
-    source_ref and the file:line citations of its prose."""
+    source_ref, the file:line citations of its prose and the file its
-    paths: set[str] = set()
+    version is read from."""
+    basis = profile.get("version_basis")
+    bases = ([basis] if isinstance(basis, dict) and "kind" in basis
+             else list(basis.values()) if isinstance(basis, dict) else [])
+    paths: set[str] = {
+        str(b["path"]) for b in bases
+        if isinstance(b, dict) and b.get("kind") == "file" and b.get("path")
+    }
     for key, value in _walk(profile):
         if key == "source_ref" and isinstance(value, (str, dict)):
             for _label, refs in source_ref_values(value):
@@ -355,6 +362,30 @@ def _commit_url(repo: upstream.Repo, pin: str) -> str:
     return f"{base}/commit/{pin}"
 
 
+def version_basis_url(profile: dict, basis: dict, mode: str = "") -> str:
+    """Where a reader checks a version: the file stating it at the pin, or
+    the tag it counts from. "" when neither can be linked.
+
+    The Version cell printed `libretro/handy.h` as bare text beside a
+    pinned Revision link, so the #define it reads was one search away.
+    """
+    field = "upstream" if basis.get("repo") == "upstream" else "source"
+    declared = next(iter(_repo_pins(profile, field, mode)), None)
+    repo = _browse_repo(declared[0]) if declared else None
+    if repo is None:
+        return ""
+    kind, path, tag = basis.get("kind"), str(basis.get("path") or ""), str(basis.get("tag") or "")
+    if kind == "file" and path:
+        pin = next((p for r, p in _forge_sources(profile, mode)
+                    if (r.host, r.slug) == (repo.host, repo.slug)), "")
+        return _source_permalink(repo, pin, path, None, None) if pin and tree_holds(
+            repo, pin, path) else ""
+    if kind == "tag" and tag and repo.family in ("github", "gitlab"):
+        tree = "/-/tree/" if repo.family == "gitlab" else "/tree/"
+        return f"https://{repo.host}/{repo.owner}/{repo.name}{tree}{urllib.parse.quote(tag)}"
+    return ""
+
+
 def revision_markdown(profile: dict) -> str:
     """Each repository the profile reads, linked at the commit it was read at.
 

scripts/sitestarlight.py

@@ -244,10 +244,21 @@ def _tooltip_line(line: str, pattern: re.Pattern, pending: dict[str, str]) -> st
     return "".join(out)
 
 
+_NAME_BEFORE = re.compile(r"(?:/|\b[A-Z][\w'-]*[ \t]+)$")
+
+
+def _in_proper_name(term: str, before: str) -> bool:
+    """A capitalised term after a capitalised word or a slash is part of a
+    name: Free Unix Spectrum Emulator, Windows Terminal, Documents/Dolphin
+    Emulator. An acronym (BIOS, ZIP) is the term wherever it stands."""
+    first = term.split()[0]
+    return first[:1].isupper() and not first.isupper() and bool(_NAME_BEFORE.search(before))
+
+
 def _tooltip_text(text: str, pattern: re.Pattern, pending: dict[str, str]) -> str:
     def swap(match: re.Match) -> str:
         term = match.group(1)
-        if term not in pending:
+        if term not in pending or _in_proper_name(term, text[:match.start()]):
             return term
         definition = pending.pop(term)
         return f'<abbr class="rb-term" title="{html.escape(definition)}">{term}</abbr>'

scripts/validate_pr.py

@@ -21,6 +21,7 @@ import argparse
 import json
 import os
 import re
+import shutil
 import subprocess
 import sys
 from pathlib import Path
@@ -36,6 +37,7 @@ from common import (
 )
 from hashing import parse_hash_list
 from nativemode import digest_algorithm
+from dedup import _in_nodedup_dir
 from validation import frontend_digest_matches, inner_rom_check
 
 try:
@@ -79,6 +81,13 @@ class ValidationResult:
         self.md5 = ""
         self.crc32 = ""
         self.size = 0
+        # "held" (a copy of a file the collection has) or "unknown" (no list
+        # or profile declares it): no check fails, and neither is a file
+        # the pull request adds to the collection.
+        self.outcome = ""
+        # Where --place copies the file: one folder and one declared name, or
+        # the .variants/ slot of a name the collection already holds.
+        self.destination = ""
 
     def add_check(self, passed: bool, message: str) -> None:
         self.checks.append(("PASS" if passed else "FAIL", message))
@@ -93,8 +102,12 @@ class ValidationResult:
     def passed(self) -> bool:
         return all(s != "FAIL" for s, _ in self.checks)
 
+    @property
+    def status(self) -> str:
+        return "FAIL" if not self.passed else "REVIEW" if self.outcome else "PASS"
+
     def to_markdown(self) -> str:
-        status = "PASS" if self.passed else "FAIL"
+        status = self.status
         lines = [f"### {status} `{self.filename}`"]
         lines.append("")
         lines.append(f"- **Path**: `{self.filepath}`")
@@ -147,7 +160,7 @@ def load_platform_hashes(platforms_dir: str, db: dict | None = None) -> dict:
     under, read from the other files of that system it holds.
     """
     known = {"sha1": set(), "md5": set(), "sha256": set(), "crc32": set(),
-             "names": set(), "folders": {}, "sizes": {}, "entries": {}}
+             "names": set(), "folders": {}, "sizes": {}, "entries": {}, "named": {}}
     systems_of: dict[str, set[str]] = {}
     folders_of: dict[str, dict[str, int]] = {}
     files = (db or {}).get("files", {})
@@ -194,6 +207,9 @@ def load_platform_hashes(platforms_dir: str, db: dict | None = None) -> dict:
                     )
                 for value in (*sha1s, *md5s):
                     systems_of.setdefault(value, set()).add(system_id)
+                for value in (*sha1s, *md5s, *sha256s):
+                    if "name" in file_entry:
+                        known["named"].setdefault(value, set()).add(file_entry["name"])
                 _note_folder(folders_of, system_id, files, by_md5, sha1s, md5s)
 
     known["folders"] = _system_homes(systems_of, folders_of)
@@ -251,6 +267,7 @@ def load_profile_hashes(
     """
     known: dict[str, dict] = {
         "sha1": {}, "sha256": {}, "md5": {}, "crc32": {}, "sizes": {}, "folders": {},
+        "named": {},
     }
     systems_of: dict[str, set[str]] = {}
     folders_of: dict[str, dict[str, int]] = {}
@@ -274,6 +291,8 @@ def load_profile_hashes(
             for algo, values in hashes.items():
                 for value in values:
                     known[algo].setdefault(value, set()).add(emulator)
+                    if isinstance(entry.get("name"), str):
+                        known["named"].setdefault(value, set()).add(entry["name"])
             # The entry's own system, or the profile's when it emulates one.
             system = entry.get("system") or (
                 declared_systems[0] if len(declared_systems) == 1 else None
@@ -374,9 +393,15 @@ def validate_file(
                 # PR that regenerated database.json was told each new file
                 # duplicated itself.
                 result.add_check(True, "Already in the database at this path")
+            elif _in_nodedup_dir(relative):
+                result.add_info(
+                    f"Identical to `{existing['path']}`, kept: this tree is read by path"
+                )
             else:
+                result.outcome = "held"
                 result.add_warning(
-                    f"Duplicate: identical file already exists at `{existing['path']}`"
+                    f"Already held at `{existing['path']}`: nothing to add, "
+                    "leave this copy out of the pull request"
                 )
         else:
             result.add_check(True, "Not a duplicate in database")
@@ -421,8 +446,11 @@ def validate_file(
             "Filename matches a known requirement but hash differs - may be a variant"
         )
     else:
+        result.outcome = "unknown"
         result.add_warning(
-            "File not referenced in any platform config - needs manual review"
+            "Not recognised: no platform list or emulator profile declares this "
+            "file. Say in the pull request which emulator reads it, for which "
+            "system, and where the dump comes from"
         )
 
     normalized = os.path.normpath(relative)
@@ -452,11 +480,27 @@ def validate_file(
             result.add_check(True, f"Correct placement: bios/{parts[1]}/{parts[2]}/")
         else:
             result.add_warning("File should be in bios/Manufacturer/Console/ structure")
+    elif result.outcome == "held":
+        pass
     elif expected_folders:
         # The contributor asks before forking: say where it goes.
+        names = {
+            name
+            for table in (platform_hashes, profile_hashes)
+            for value in (result.sha1, result.md5, hashes.get("sha256", ""))
+            for name in table.get("named", {}).get(value, ())
+        }
+        name = next(iter(names)) if len(names) == 1 else ""
+        if len(expected_folders) == 1 and name and "/" not in name:
+            folder = expected_folders[0]
+            result.destination = (
+                f"{folder}.variants/{name}.{result.sha1[:8]}"
+                if os.path.exists(folder + name) else folder + name
+            )
         result.add_warning(
             "File is not under bios/ yet. Its system's files are in "
             + " or ".join(f"`{f}`" for f in expected_folders)
+            + (f". --place copies it to `{result.destination}`" if result.destination else "")
         )
     elif _variant_home(db, result.filename, result.sha1):
         result.add_warning(
@@ -485,9 +529,11 @@ def validate_file(
     if (name_known and not (sha1_known or md5_known or other_known or accepted_by)
             and not by_profile and size_fits):
         home = _variant_home(db, result.filename)
+        slot = f"{home}.variants/{result.filename}.{result.sha1[:8]}"
+        if home and not normalized.startswith("bios" + os.sep):
+            result.destination = slot
         result.add_info(
-            "This may be a valid variant. If accepted, it will be placed in "
+            f"This may be a valid variant. If accepted, it will be placed in `{slot}`"
-            f"`{home}.variants/{result.filename}.{result.sha1[:8]}`"
         )
 
     return result
@@ -558,30 +604,35 @@ def _repo_key(url: str) -> str:
 
 
 def _project_main_line() -> str | None:
-    """The project's main line, fetched when no remote of this clone names it.
+    """The project's main line, fetched when this clone does not hold it.
 
     A fork's clone has origin alone. Once the commit is pushed to the fork's
     main, origin/main is HEAD and the diff against it was empty, so --changed
-    found nothing and exited 0. None leaves the upstream/ and origin/ lines
+    found nothing and exited 0. A remote added as CONTRIBUTING says and never
-    in charge: a remote of that name names the project, or the clone has no
+    fetched did the same: upstream/main did not exist and origin/main won.
-    remote at all and only its uncommitted files can be read.
+    None leaves the upstream/ and origin/ lines in charge when the clone has
+    no remote at all and only its uncommitted files can be read.
     """
     remotes = _git_lines("remote") or []
     if not remotes:
         return None
     project = {_repo_key(url) for url in PROJECT_URLS}
+    source = PROJECT_URLS[0]
     for remote in remotes:
         urls = _git_lines("remote", "get-url", "--all", remote) or []
         if any(_repo_key(url) in project for url in urls):
-            return None if remote in ("upstream", "origin") else f"{remote}/main"
+            if _git_lines("rev-parse", "--verify", "--quiet", f"{remote}/main"):
+                return f"{remote}/main"
+            source = remote
+            break
     fetched = subprocess.run(
-        ["git", "fetch", "--quiet", "--no-tags", PROJECT_URLS[0], "main"],
+        ["git", "fetch", "--quiet", "--no-tags", source, "main"],
         capture_output=True, text=True, check=False,
     )
     if fetched.returncode != 0:
         raise RuntimeError(
-            f"no remote of this clone is {PROJECT_URLS[0]}, and fetching its "
+            f"this clone does not hold the main of {PROJECT_URLS[0]}, and "
-            f"main failed: {fetched.stderr.strip()}"
+            f"fetching its main from {source} failed: {fetched.stderr.strip()}"
         )
     return "FETCH_HEAD"
 
@@ -675,6 +726,28 @@ def _files_to_validate(
     return files
 
 
+def place(results: list[ValidationResult]) -> list[str]:
+    """Copy each placeable file to its destination, never over a file.
+
+    The report named the folder of each file and left 40 copies from a card
+    to be made by hand, each .variants/ name retyped. The source is copied,
+    not moved: it is often a card or another tree the contributor keeps.
+    """
+    placed = []
+    for r in results:
+        if not (r.passed and r.destination and not r.outcome):
+            continue
+        target = Path(r.destination)
+        if target.exists():
+            print(f"not placed: {target} exists", file=sys.stderr)
+            continue
+        target.parent.mkdir(parents=True, exist_ok=True)
+        shutil.copy2(r.filepath, target)
+        placed.append(r.destination)
+        print(f"placed {r.filepath} at {r.destination}", file=sys.stderr)
+    return placed
+
+
 def _load_database_if_any(path: str) -> dict | None:
     """The database, optional only where the default is meant.
 
@@ -709,6 +782,11 @@ def main() -> None:
         "--markdown", action="store_true", help="Output as markdown (for PR comments)"
     )
     output.add_argument("--json", action="store_true", help="Output as JSON")
+    parser.add_argument(
+        "--place", action="store_true",
+        help="copy each recognised file from outside bios/ to the folder and name "
+        "the report gives, or to .variants/ beside the name it varies",
+    )
     args = parser.parse_args()
 
     # The database, platforms and profiles are read from the repository
@@ -745,6 +823,8 @@ def main() -> None:
                 {
                     "file": r.filepath,
                     "passed": r.passed,
+                    "outcome": r.outcome or "new",
+                    "destination": r.destination,
                     "sha1": r.sha1,
                     "md5": r.md5,
                     "size": r.size,
@@ -754,7 +834,15 @@ def main() -> None:
         print(json.dumps(output, indent=2))
     elif args.markdown:
         lines = ["## BIOS Validation Report", ""]
-        status = "all checks passed" if all_passed else "some checks failed"
+        held = sum(r.outcome == "held" for r in results)
+        unknown = sum(r.outcome == "unknown" for r in results)
+        review = ", ".join(part for part in (
+            f"{held} already held" if held else "",
+            f"{unknown} not recognised" if unknown else "",
+        ) if part)
+        status = ("some checks failed" if not all_passed
+                  else f"no check failed, to review: {review}" if review
+                  else "all checks passed")
         lines.append(f"**Status**: {status}")
         lines.append("")
 
@@ -765,8 +853,7 @@ def main() -> None:
         print("\n".join(lines))
     else:
         for r in results:
-            status = "PASS" if r.passed else "FAIL"
+            print(f"\n[{r.status}] {r.filepath}")
-            print(f"\n[{status}] {r.filepath}")
             if r.sha1:
                 print(f"  SHA1: {r.sha1}")
                 print(f"  MD5:  {r.md5}")
@@ -783,6 +870,9 @@ def main() -> None:
                 )
                 print(f"  [{marker}] {m}")
 
+    if args.place:
+        place(results)
+
     if not all_passed:
         sys.exit(1)
 

tests/test_audit_regressions.py

@@ -6781,7 +6781,8 @@ class SkipPacksHelpSaysWhatRuns(unittest.TestCase):
         )
         text = " ".join(proc.stdout.split())
         self.assertNotIn("Only regenerate DB and verify", text)
-        self.assertIn("Steps 7 and 8 still rewrite README.md and the site pages in website/", text)
+        self.assertIn("step 4e still rewrites install/games/, and steps 7 to 8b rewrite README.md "
+                      "and the site unless --skip-docs is given", text)
         self.assertNotIn("docs/", text)
 
     def test_every_step_it_names_is_one_it_skips(self):

tests/test_contributing_text.py

@@ -78,6 +78,36 @@ class BeforeOpeningAPullRequest(unittest.TestCase):
         self.assertIn("profile_sync.py --added --base upstream/main --strict", block)
         self.assertNotIn("--changed --strict", block)
 
+    def test_it_names_the_site_build_and_the_packages_ci_runs(self):
+        """A wiki-only change passed every listed check and failed CI on
+        validate_site.py, which the list did not name."""
+        text = generate_readme.generate_contributing()
+        section = text.split("## Before opening a PR", 1)[1].split("\n## ", 1)[0]
+        workflow = (REPO_ROOT / ".github" / "workflows" / "validate.yml").read_text(encoding="utf-8")
+        for command in ("npm ci --prefix website", "npm run build --prefix website",
+                        "validate_site.py --site-dir website/dist",
+                        "vermin --no-tips -t=3.8- --violations install.py"):
+            with self.subTest(command=command):
+                self.assertIn(command, workflow)
+                self.assertIn(command, section)
+        installed = re.search(r'^\s*run: pip install (pyyaml "jsonschema.*markdown.*)$',
+                              workflow, re.M).group(1)
+        pip_line = next(line for line in section.splitlines() if line.startswith("pip install"))
+        for package in (re.split(r"[\[=]", token.strip('"'))[0] for token in installed.split()):
+            with self.subTest(package=package):
+                self.assertRegex(pip_line, rf"\b{package}\b")
+
+    def test_a_file_too_large_for_git_has_a_route(self):
+        """The guide sent any file over 25 MB through a clone, which GitHub
+        refuses past 100 MB, and only a maintainer can write to the release."""
+        import validate_pr
+
+        text = generate_readme.generate_contributing()
+        section = text.split("## What happens next", 1)[1].split("\n## ", 1)[0]
+        self.assertEqual(validate_pr.MAX_FILE_SIZE, 100 * 1024 * 1024)
+        self.assertIn("100 MB", section)
+        self.assertIn(f"]({validate_pr.LARGE_FILE_ISSUE_URL})", section)
+
     def test_it_installs_what_the_checks_need_and_no_site_package(self):
         """The site is built by Node from website/package-lock.json: a pip
         line naming the old theme installed packages nothing reads."""

tests/test_deterministic_zip.py

@@ -252,6 +252,21 @@ class TestPackDeterminism(unittest.TestCase):
             names = {i.filename.rsplit("/", 1)[-1] for i in zf.infolist()}
         self.assertIn("romset.zip", names)
 
+    def test_a_romset_is_the_collection_bytes_and_the_wiki_says_so(self):
+        """The pack copies each ZIP byte for byte since fed03c3db, and the
+        wiki still told a third party to rebuild romsets to compare."""
+        pack = self._build("out1")
+        with zipfile.ZipFile(pack) as zf:
+            member = next(n for n in zf.namelist() if n.endswith("romset.zip"))
+            self.assertEqual(zf.read(member), (self.bios / "romset.zip").read_bytes())
+        root = Path(__file__).resolve().parent.parent / "wiki"
+        for page in ("architecture.md", "tools.md"):
+            with self.subTest(page=page):
+                text = " ".join((root / page).read_text(encoding="utf-8").split())
+                self.assertNotIn("rebuilt deterministically", text)
+                self.assertNotIn("rebuilds MAME", text)
+                self.assertNotIn("Rebuild MAME BIOS ZIPs", text)
+
     def test_every_member_carries_the_fixed_epoch(self):
         """A member's date must come from the build's inputs, not its clock.
 

tests/test_doc_pages.py

@@ -1799,5 +1799,91 @@ class ReportExcerptIsTheRun(unittest.TestCase):
                 self.assertIn(line, printed)
 
 
+class ClaimRuleNamesTheSizeProof(unittest.TestCase):
+    """architecture.md said a file found by its name claims nothing, while
+    slots.py has counted a name-found copy at the declared size against a
+    platform file of another size since 2b535525e."""
+
+    def test_the_page_states_what_slots_counts(self):
+        import slots
+
+        self.assertIn(slots.SIZE_EXACT, slots.PROVEN_STATUSES)
+        page = " ".join((ROOT / "wiki" / "architecture.md").read_text(encoding="utf-8").split())
+        section = page.split("### When both layers claim one destination", 1)[1].split("###", 1)[0]
+        self.assertIn(f"(`{slots.SIZE_EXACT}`)", section)
+        self.assertIn("Two files of equal size are revisions", section)
+
+
+class CrossReferenceSettlesBySlotAndSize(unittest.TestCase):
+    """reading-reports.md said the cross-reference checks whether a profile's
+    files "appear in the platform config", a by-name rule, while verify
+    settles a file only at its slot or at a size the core accepts."""
+
+    def test_the_page_states_the_rule_verify_reads(self):
+        from validation import declaration_settles
+
+        rom1 = {"name": "ROM1.BIN", "size": 32768, "validation": ["size"]}
+        self.assertFalse(declaration_settles(rom1, "ROM1.BIN", {"ROM1.BIN": [(4096, "rom1.bin")]},
+                                             dest="dosbox/ROM1.BIN"))
+        self.assertTrue(declaration_settles(rom1, "ROM1.BIN", {"ROM1.BIN": [(32768, "rom1.bin")]},
+                                            dest="dosbox/ROM1.BIN"))
+        page = " ".join((ROOT / "wiki" / "reading-reports.md").read_text(encoding="utf-8").split())
+        section = page.split("### Cross-reference section", 1)[1].split("###", 1)[0]
+        self.assertIn("declaration_settles()", section)
+        self.assertIn("A name alone is not enough", section)
+        self.assertNotIn("appear in the platform config", section)
+
+
+class CommittedLargeFilesAreTheOnesNamed(unittest.TestCase):
+    """release-process.md named tbblue.mmc among the committed files over
+    50 MB two weeks after it left bios/, and not uqm-remix-disc1.uqm."""
+
+    def test_the_list_is_the_tree(self):
+        import subprocess
+
+        listed = subprocess.run(["git", "ls-files", "-z", "bios"], cwd=ROOT, capture_output=True,
+                                check=True).stdout.decode("utf-8").split("\0")
+        large = sorted(Path(path).name for path in listed
+                       if path and (ROOT / path).is_file()
+                       and (ROOT / path).stat().st_size > 50_000_000)
+        page = " ".join((ROOT / "wiki" / "release-process.md").read_text(encoding="utf-8").split())
+        said = re.search(r"Size does not decide it: (\w+) committed files are over 50 MB \(([^)]*)\)",
+                         page)
+        self.assertIsNotNone(said)
+        self.assertEqual(sorted(re.findall(r"`([^`]+)`", said.group(2))), large)
+        self.assertEqual(said.group(1), {3: "three", 4: "four", 5: "five"}.get(len(large)))
+
+
+class SkipFlagsNameTheSteps(unittest.TestCase):
+    """--skip-packs said "add --skip-docs to stop after verify" while 4d and
+    4e still ran, and the flag table left 8b out of --skip-docs."""
+
+    STEP = r"\b\d[a-z]?\d?\b"
+
+    def test_the_flag_row_the_help_and_the_usage_agree_with_the_step_table(self):
+        page = (ROOT / "wiki" / "tools.md").read_text(encoding="utf-8")
+        table = page.split("Pipeline steps:", 1)[1].split("Pipeline flags:", 1)[0]
+        source = (ROOT / "scripts" / "pipeline.py").read_text(encoding="utf-8")
+        helps = {}
+        for node in ast.walk(ast.parse(source)):
+            if (isinstance(node, ast.Call) and getattr(node.func, "attr", "") == "add_argument"
+                    and node.args and isinstance(node.args[0], ast.Constant)):
+                helps.update({node.args[0].value: ast.literal_eval(kw.value)
+                              for kw in node.keywords if kw.arg == "help"})
+        for flag in ("--skip-packs", "--skip-docs"):
+            steps = set()
+            for row in table.splitlines():
+                cells = [c.strip() for c in row.split("|")]
+                if len(cells) > 3 and cells[3].startswith(f"`{flag}`"):
+                    steps.add(cells[1])
+            self.assertTrue(steps)
+            row = re.search(rf"^\| `{flag}` \| skip steps ([^.|]*)", page, re.M).group(1)
+            usage = re.search(rf"pipeline\.py {flag} +# skip steps (.*)$", source, re.M).group(1)
+            said = helps[flag].split(":", 1)[0]
+            for text in (row, usage, said):
+                with self.subTest(flag=flag, text=text):
+                    self.assertEqual(set(re.findall(self.STEP, text)), steps)
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_file_consumers.py

@@ -17,7 +17,7 @@ from pathlib import Path
 REPO_ROOT = Path(__file__).resolve().parent.parent
 sys.path.insert(0, str(REPO_ROOT / "scripts"))
 
-from cross_reference import FileConsumers  # noqa: E402
+from cross_reference import FileConsumers, entry_target  # noqa: E402
 from generate_site import generate_system_page  # noqa: E402
 
 _TMP = tempfile.TemporaryDirectory()
@@ -246,7 +246,6 @@ class EveryReaderIsTheEmulatorPackFile(unittest.TestCase):
             build_zip_contents_index, load_data_dir_registry,
             load_emulator_profiles, resolve_local_file,
         )
-        from validation import with_archive_members
 
         db_path = REPO_ROOT / "database.json"
         if not db_path.exists():
@@ -275,9 +274,12 @@ class EveryReaderIsTheEmulatorPackFile(unittest.TestCase):
                 if members:
                     from_trees |= {(sha1, name) for sha1 in members.values()}
                     continue
+                # A ROM read inside an archive is the archive, as the
+                # emulator pack resolves it (archive_entry).
+                target, dest = entry_target(entry, files)
                 path, status = resolve_local_file(
-                    {**with_archive_members(entry, files), "source_profile": name},
+                    {**target, "source_profile": name},
-                    db, zips, dest_hint=entry.get("path") or "",
+                    db, zips, dest_hint=dest,
                     data_dir_registry=registry,
                 )
                 if path and status not in ("not_found", "hash_mismatch"):

tests/test_profile_sync.py

@@ -5974,6 +5974,13 @@ class TestBranchProfiles(unittest.TestCase):
             ["draft", "new"],
         )
 
+    def test_an_unfetched_base_says_to_fetch(self):
+        """`--base upstream/main` with upstream added and never fetched
+        stopped on git's "Not a valid object name" and no next step."""
+        with self.assertRaises(SystemExit) as raised:
+            profile_sync.branch_profiles("emulators", "upstream/main", added_only=True)
+        self.assertIn("git fetch upstream", str(raised.exception))
+
     def test_the_flag_selects_those_profiles(self):
         args = profile_sync.build_parser().parse_args(["--changed", "--base", "main"])
         profiles = {"kept": {}, "edited": {}, "new": {}, "draft": {}}

tests/test_prose_style.py

@@ -308,6 +308,22 @@ class ProfilesStateFactsAboutOtherProjects(unittest.TestCase):
         ]
         self.assertEqual(offenders, [], "\n".join(offenders[:40]))
 
+    # A project's state is a dated fact or its own words, never a verdict:
+    # "Upstream project is dead" (yabause), "abandoned fork" (gearboy).
+    PROJECT_VERDICTS = re.compile(
+        r"\b(?:(?:project|upstream)(?:\s+project)?\s+(?:is\s+)?dead|"
+        r"abandoned\s+(?:fork|since)|project\s+abandoned|never-completed)\b",
+        re.IGNORECASE,
+    )
+
+    def test_no_verdict_on_a_project(self):
+        offenders = [
+            f"{path.relative_to(ROOT)}: {found.group(0)}"
+            for path in sorted((ROOT / "emulators").glob("*.yml"))
+            for found in self.PROJECT_VERDICTS.finditer(path.read_text(encoding="utf-8"))
+        ]
+        self.assertEqual(offenders, [], "\n".join(offenders[:40]))
+
     def test_no_grade_on_another_projects_file(self):
         offenders = [
             f"{path.relative_to(ROOT)}: {line.strip()[:100]}"

tests/test_refresh_failures.py

@@ -130,7 +130,7 @@ class AForkPushedToItsOwnMain(unittest.TestCase):
             # A project that cannot be reached is an error, never "nothing".
             missing = self._changed(clone, str(Path(tmp, "absent")))
         self.assertNotEqual(missing.returncode, 0)
-        self.assertIn("fetching its main failed", missing.stderr)
+        self.assertIn("fetching its main from", missing.stderr)
 
 
 class RefreshFailuresAreFailures(unittest.TestCase):

tests/test_site_components.py

@@ -334,6 +334,22 @@ class NoLevelBadgeOnATabRoot(unittest.TestCase):
                 self.assertNotIn('class="rb-level ', _title_through_heading(_built(page)))
 
 
+class OneTabIsNoTabBar(unittest.TestCase):
+    """142 tab bars held one tab, most of them "On the device" right under
+    the heading "On the device"."""
+
+    def test_a_single_item_stands_alone(self):
+        import sitemarkup
+        import siteguides
+
+        self.assertEqual(sitemarkup.tabs([("Linux", ["run it"])]), ["**Linux**", "", "run it", ""])
+        self.assertEqual(sitemarkup.tabs([("", ["run it"])]), ["run it", ""])
+        self.assertIn('<Tabs syncKey="os">', sitemarkup.tabs([("A", ["a"]), ("B", ["b"])]))
+        device = "\n".join(siteguides.terminal_tabs(("device",), {}))
+        self.assertNotIn("<Tab", device)
+        self.assertNotIn("On the device", device)
+
+
 class FooterLicenceMatchesTheLicenceFile(unittest.TestCase):
     """The footer said MIT covers the tooling. LICENSE names the metadata too."""
 
@@ -353,15 +369,24 @@ class WideTablesStackAndFiltersAreBuilt(unittest.TestCase):
     def test_every_wide_table_stacks(self):
         for page in ("emulators/duckstation", "gaps", "gaps-files-to-source", "emulators", "data",
                      "wiki/verification-modes"):
-            html = (DIST / page / "index.html")
+            for table in _tables(_built(page)):
-            if not html.is_file():
-                continue
-            for table in _tables(_read(html)):
                 if len(_headers(table)) < 4:
                     continue
                 with self.subTest(page=page, header=_headers(table)):
                     self.assertIn('data-rb-stack="true"', table.split(">", 1)[0])
 
+    def test_no_build_is_a_skip_not_a_pass(self):
+        """Every missing page was a `continue`: with no build the test ran no
+        assertion and still reported ok."""
+        from unittest import mock
+
+        module = sys.modules[__name__]
+        case = WideTablesStackAndFiltersAreBuilt("test_every_wide_table_stacks")
+        result = unittest.TestResult()
+        with mock.patch.object(module, "DIST", REPO_ROOT / "tmp" / "no-such-dist"):
+            case.run(result)
+        self.assertEqual(len(result.skipped), 1)
+
     def test_long_tables_get_their_labels_at_runtime(self):
         """Labels on every cell of a long stacked table were 122 KB of
         emulators/mame and took it over the 600,000-byte page budget."""

tests/test_site_emulator_page.py

@@ -68,6 +68,38 @@ class WhoShipsTheBuildAndWhy(unittest.TestCase):
                       "(https://example.org/news) (version 13.0 announced there, "
                       "found only as rebuilt copies) |", page)
 
+    def test_the_version_links_its_proof(self):
+        """`0.97, stated in libretro/handy.h` was bare text beside a pinned
+        Revision link: the #define it reads was one search away."""
+        import generate_site as gs
+        import sitesources
+        import yaml
+
+        handy = yaml.safe_load((REPO_ROOT / "emulators" / "handy.yml").read_text(encoding="utf-8"))
+        path = handy["version_basis"]["path"]
+        self.assertIn(path, sitesources.cited_paths(handy))
+        self.assertIn(f"stated in [`{path}`](https://github.com/libretro/libretro-handy/blob/"
+                      f"{handy['source_commit']}/{path})", gs.version_markdown(handy))
+        tagged = {"source": "https://github.com/owner/emu", "source_commit": "a" * 40,
+                  "core_version": "1.2-3-gaaaaaaa",
+                  "version_basis": {"kind": "tag", "tag": "1.2", "distance": 3}}
+        self.assertEqual(gs.version_markdown(tagged), "1.2-3-gaaaaaaa, 3 commits after tag "
+                         "[`1.2`](https://github.com/owner/emu/tree/1.2)")
+        # A file no recorded tree confirms stays plain: no link beats a 404.
+        unread = {**tagged, "version_basis": {"kind": "file", "path": "src/none.h"}}
+        self.assertEqual(gs.version_markdown(unread), "1.2-3-gaaaaaaa, stated in `src/none.h`")
+
+    def test_the_upstream_pin_says_why_it_is_held(self):
+        """mupen64plus showed its core revision and not that AmberELEC
+        builds that one: read as a profile behind HEAD."""
+        page = _page("mupen64plus", {
+            "type": "libretro",
+            "upstream_frozen": "AmberELEC builds the core at this revision.",
+            "files": [],
+        })
+        self.assertIn("| Upstream revision held | AmberELEC builds the core at this "
+                      "revision. |", page)
+
 
 class ACardKeepsItsAddress(unittest.TestCase):
     """emulators/raze/#file-duke3d-grp stopped pointing at its card once a

tests/test_site_emulators.py

@@ -254,11 +254,16 @@ class WhatTheAppShipsIsNoFileToGet(unittest.TestCase):
         self.assertIn(">comes with the app</span>", badges)
         self.assertNotIn(">required</span>", badges)
         self.assertNotIn(">missing</span>", badges)
-        # Installed by one build only: the other still reads it from the folder.
+        # Narrowed to one build, the entry is that build's alone, as verify
+        # and the pack read it: EightyOne installs its ROMs, and the libretro
+        # core, which embeds them, reads no entry. Nothing to fetch.
         dual = " ".join(gs._file_badges({"name": "a", "bundled": True, "mode": "standalone",
                                          "required": True}, False))
-        self.assertIn(">required</span>", dual)
+        self.assertNotIn(">required</span>", dual)
+        self.assertNotIn(">missing</span>", dual)
         self.assertIn("installed with the standalone emulator", dual)
+        rom = next(f for f in _profile("81")["files"] if f["name"] == "jupiterace.rom")
+        self.assertTrue(gs.comes_with_app(rom))
 
     def test_counts_leave_them_out(self):
         shipped = sum(1 for f in self.profile["files"] if gs.comes_with_app(f))
@@ -399,7 +404,9 @@ class AnAppThatReadsAnyBios(unittest.TestCase):
         install = self.page[self.page.index("Install this emulator's files"):]
         install = install[:install.index("</details>")]
         self.assertRegex(install, r"With [^:]+: the installer finds")
-        self.assertEqual(set(re.findall(r'<TabItem label="([^"]+)">', install)), {"On the device"})
+        # One place to run it, so no tab bar (one tab offers no choice).
+        self.assertNotIn("<Tabs", install)
+        self.assertIn("curl -fsSL", install)
         nether = _pages("nethersx2")["nethersx2"]
         if "--emulator nethersx2" in nether:
             self.assertIn("With ", nether)

tests/test_site_entry.py

@@ -443,14 +443,22 @@ class PackNamesItsVersion(unittest.TestCase):
         block = generate_site._pack_buttons("batocera")
         version = generate_site._pack_version(release_record.pack_name_for("batocera", record))
         self.assertIn(f"Made for Batocera {version}. "
-                      + generate_site.OTHER_VERSION, block)
+                      + generate_site.other_version("batocera"), block)
 
     def test_a_version_gap_is_said_once(self):
         record = {"tag": "v2026.10.10", "packs": {"RetroBat_1.0_BIOS_Pack.zip": {
             "assets": ["RetroBat_1.0_BIOS_Pack.zip"], "platforms": ["retrobat"]}}}
         released = generate_site.release_line("retrobat", record)
         line = generate_site.version_line("retrobat", record, named=len(released) > 1)
-        self.assertEqual(line, generate_site.OTHER_VERSION)
+        self.assertEqual(line, generate_site.other_version("retrobat"))
+
+    def test_another_version_is_sent_to_the_check_commands(self):
+        """A ZIP reader read "Run `--check` after copying." with no command."""
+        line = generate_site.other_version("emudeck")
+        self.assertNotIn("Run `--check`", line)
+        self.assertIn("](../platforms/emudeck.md#how-to-check)", line)
+        page = _page("platforms/emudeck.md")
+        self.assertIn("{ #how-to-check }", page)
 
     def test_the_v_prefix_is_dropped(self):
         self.assertEqual(generate_site.setup_version("retroarch", {}),
@@ -925,7 +933,9 @@ class ARouteWithoutANetwork(unittest.TestCase):
         self.assertEqual(siteguides.GUIDES["retropie"].card_run, ("linux",))
         text = _plain_page("platforms/retropie.md")
         card = text.split("No network?", 1)[1].split("## On the device", 1)[0]
-        self.assertEqual(_tab_labels(card), {"Linux"})
+        # One system, so no tab bar: its label leads the command.
+        self.assertEqual(_tab_labels(card), set())
+        self.assertIn("**Linux**", card)
 
     def test_no_zip_is_promised_for_an_unmarked_setup(self):
         for name in SETUPS:

tests/test_site_reference.py

@@ -422,7 +422,11 @@ class BuiltBatoceraPage(unittest.TestCase):
         # PlayStation's list is long enough to sit on a page of its own.
         pages = [main] + [p.read_text(encoding="utf-8") for p in
                           (DIST / "platforms").glob("batocera-sony-playstation*/index.html")]
-        html = "".join(pages)
+        # Read as a browser shows it: a page with no tab bar is plain
+        # Markdown, which keeps the entity a name's underscore is written as.
+        import html as _html
+
+        html = _html.unescape("".join(pages))
         for text in ("Served by:", "scph1001_v20.bin", "https://github.com/stenzek/duckstation/blob/"):
             with self.subTest(text=text):
                 self.assertTrue(text in html)
@@ -696,6 +700,23 @@ class ManifestsArePublished(unittest.TestCase):
         self.assertLess(page.index("## Stable JSON endpoints"), page.index("The DAT is Logiqx XML"))
 
 
+class TheLookupRecipeSaysWhatASha1Misses(unittest.TestCase):
+    """The page said a SHA1 lookup was enough, "a setup that lists a file
+    by MD5 or CRC32 is filed under the SHA1 of the collection file it names
+    too". A record whose value no collection file answers has no SHA1: 702
+    platform MD5s were out of the recipe's reach."""
+
+    def test_the_page_and_the_records_agree(self):
+        record = {"layer": "platform", "algorithm": "md5", "value": "0" * 32,
+                  "name": "dmg_boot.bin", "owner": "recalbox"}
+        self.assertEqual(generate_site._held_under_every_hash([record], {}, {}), [])
+        self.assertNotIn("sha1", record)
+        page = " ".join(generate_site.generate_data_page([]).split())
+        self.assertNotIn("none is needed", page)
+        self.assertIn("A list naming a file the collection lacks is filed under its MD5 "
+                      "or CRC32 alone", page)
+
+
 class AnIdLeadsToItsData(unittest.TestCase):
     """8.6: a system by its id, a core by the name the buildbot gives it."""
 

tests/test_site_render.py

@@ -427,5 +427,81 @@ class TheSearchSkipsLinkLists(unittest.TestCase):
         self.assertIn(short, text)
 
 
+class OneDefinitionPerName(unittest.TestCase):
+    """generate_site defined _REGION_WORDS twice: the second table replaced
+    the first at import, and _region_word, written beside the first, never
+    read it."""
+
+    def test_no_module_name_is_bound_twice(self):
+        import ast
+
+        for path in sorted((REPO_ROOT / "scripts").glob("*site*.py")):
+            seen: dict[str, int] = {}
+            for node in ast.parse(path.read_text(encoding="utf-8")).body:
+                if isinstance(node, (ast.FunctionDef, ast.ClassDef)):
+                    names = [node.name]
+                elif isinstance(node, ast.Assign):
+                    names = [t.id for t in node.targets if isinstance(t, ast.Name)]
+                elif isinstance(node, ast.AnnAssign) and isinstance(node.target, ast.Name):
+                    names = [node.target.id]
+                else:
+                    continue
+                for name in names:
+                    self.assertNotIn(name, seen, f"{path.name}:{node.lineno} binds {name} "
+                                     f"again, first at line {seen.get(name)}")
+                    seen[name] = node.lineno
+
+
+@unittest.skipUnless((REPO_ROOT / "database.json").is_file(), "database.json is not built")
+class AMemberIsReadInItsArchive(unittest.TestCase):
+    """FBNeo reads sm1.sm1 inside neogeo.zip. Resolved by its crc32, the
+    entry credited a loose sm1.sm1 FBNeo never opens, and no neogeo.zip card
+    named FBNeo, while the pack ships it one."""
+
+    def test_the_archive_is_credited_and_named_on_the_card(self):
+        import json
+
+        import generate_site as gs
+        from common import archive_entry, resolve_local_file
+        from cross_reference import FileConsumers
+
+        db = json.loads((REPO_ROOT / "database.json").read_text(encoding="utf-8"))
+        profile = yaml.safe_load((REPO_ROOT / "emulators" / "fbneo.yml").read_text(encoding="utf-8"))
+        files = profile["files"]
+        consumers = FileConsumers(db)
+        consumers.add_emulators({"fbneo": profile})
+        credited = {consumers.path_of(sha1) for sha1 in consumers.emulators}
+        shipped, _status = resolve_local_file(
+            {**archive_entry("neogeo.zip", files), "source_profile": "fbneo"}, db)
+        self.assertIn(shipped, credited)
+        self.assertFalse([p for p in credited if p.endswith("/sm1.sm1")])
+        sm1 = next(f for f in files if f["name"] == "sm1.sm1")
+        rows = "\n".join(gs._render_emulator_file(sm1, profile, consumers, files,
+                                                  lambda _f: True, owner="fbneo"))
+        self.assertIn("Collection copy: member `sm1.sm1` of", rows)
+        self.assertIn(shipped, rows)
+
+
+class NoTooltipInsideAName(unittest.TestCase):
+    """The glossary's capitalised terms matched almost only product names:
+    Free Unix Spectrum Emulator, Windows Terminal, Documents/Dolphin Emulator."""
+
+    def test_names_are_left_and_terms_are_marked(self):
+        import sitestarlight
+
+        terms = {"Emulator": "An app that acts like a console.",
+                 "Terminal": "A window where you type commands.",
+                 "BIOS": "The console's built-in software."}
+        for line in ("Fuse (Free Unix Spectrum Emulator) reads it.",
+                     "Open Windows Terminal (WSL).",
+                     "Copy it to Documents/Dolphin Emulator/Sys."):
+            with self.subTest(line=line):
+                self.assertNotIn("rb-term", sitestarlight.add_tooltips(line, terms))
+        marked = sitestarlight.add_tooltips(
+            "Emulator settings come first. The PlayStation BIOS goes there.", terms)
+        self.assertIn('title="An app that acts like a console.">Emulator</abbr>', marked)
+        self.assertIn('>BIOS</abbr>', marked)
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_site_search.py

@@ -231,6 +231,40 @@ class Descriptions(unittest.TestCase):
         self.assertTrue(text.startswith("DuckStation emulates Sony PlayStation."))
         self.assertNotIn("source-pinned", text)
 
+    def test_a_page_whose_files_come_with_the_app_promises_none(self) -> None:
+        """DaedalusX64's three files come with the app. The result read
+        "DaedalusX64 BIOS files" over "DaedalusX64 emulates Nintendo 64."."""
+        saved, shipped = dict(gs.PAGE_DATA), set(gs.NOTHING_TO_FETCH)
+        self.addCleanup(lambda: (gs.PAGE_DATA.clear(), gs.PAGE_DATA.update(saved),
+                                 gs.NOTHING_TO_FETCH.clear(), gs.NOTHING_TO_FETCH.update(shipped)))
+        with open(REPO_ROOT / "emulators" / "daedalusx64.yml", encoding="utf-8") as handle:
+            gs.record_emulator_page("daedalusx64", yaml.safe_load(handle))
+        key = Path("emulators/daedalusx64.md")
+        self.assertEqual(gs._browser_title(key, "DaedalusX64"), "DaedalusX64 emulator")
+        body = "# DaedalusX64\n\nDaedalusX64 emulates Nintendo 64.\n"
+        self.assertEqual(gs._page_description(body, key, "DaedalusX64"),
+                         "DaedalusX64 emulates Nintendo 64. Every file it reads comes with the app.")
+
+    def test_an_initial_does_not_end_the_sentence(self) -> None:
+        """hcl's description was "Hydra Castle Labyrinth: open-source port of E."."""
+        body = ("# Hydra Castle Labyrinth\n\n<details>\n<summary>Technical notes</summary>\n\n"
+                "Open-source port of E. Hashimoto's 2008 freeware Metroidvania. It reads "
+                "its data folder.\n\n</details>\n")
+        text = gs._emulator_description(body, "emulators/hcl.md", "Hydra Castle Labyrinth", "")
+        self.assertTrue(text.startswith("Hydra Castle Labyrinth: open-source port of E. "
+                                        "Hashimoto's 2008 freeware Metroidvania."), text)
+        self.assertNotIn("It reads", text)
+
+    def test_a_platform_page_says_what_it_gives(self) -> None:
+        """All 18 ended on the label of a link a snippet cannot follow,
+        "ROCKNIX website.", and never named the BIOS files or their folder."""
+        body = ("# ROCKNIX\n\nA system for handhelds. [ROCKNIX website](https://rocknix.org).\n\n"
+                "## On the device { #install }\n")
+        self.assertEqual(
+            gs._page_description(body, Path("platforms/rocknix.md"), "ROCKNIX"),
+            "The BIOS files for ROCKNIX, the folder they go in and the command that installs "
+            "them. A system for handhelds.")
+
     def test_a_console_page_is_described_by_its_start_line(self) -> None:
         body = "# Sega: Dreamcast\n\nPart of Sega.\n\n**Start with**: `dc_boot.bin`, `dc_flash.bin`.\n"
         text = gs._page_description(body, Path("systems/sega-dreamcast.md"), "Sega: Dreamcast")

tests/test_site_system_names.py

@@ -116,5 +116,24 @@ class SystemNames(unittest.TestCase):
         self.assertFalse(names & siterender.program_names(self.profiles))
 
 
+class SectionsComeBeforeThePlan(unittest.TestCase):
+    """plan_emulator_pages names systems. Run before set_console_sections,
+    it made siterender read ./database.json and ./emulators whatever --db
+    and --emulators-dir gave, so the parts were cut on another collection."""
+
+    def test_generate_installs_the_sections_before_planning(self):
+        import ast
+        import inspect
+
+        import generate_site
+
+        tree = ast.parse(inspect.getsource(generate_site._generate))
+        lines = {}
+        for node in ast.walk(tree):
+            if isinstance(node, ast.Call) and isinstance(node.func, ast.Name):
+                lines.setdefault(node.func.id, node.lineno)
+        self.assertLess(lines["set_console_sections"], lines["plan_emulator_pages"])
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_validate_pr.py

@@ -139,9 +139,25 @@ class AFileOnlyACoreReadsIsKnown(_Fixture):
         )
         self.assertIn(
             ("WARN", "File is not under bios/ yet. Its system's files are in "
-                     "`bios/Maker/Console/`"),
+                     "`bios/Maker/Console/`. --place copies it to `bios/Maker/Console/b.bin`"),
             result.checks,
         )
+        self.assertEqual(result.destination, "bios/Maker/Console/b.bin")
+        # A file of that name already there: the copy is a variant.
+        self._bios_file("bios/Maker/Console/b.bin", b"another")
+        self.assertEqual(
+            validate_pr.validate_file("Downloads/b.bin", db, EMPTY_HASHES, known).destination,
+            f"bios/Maker/Console/.variants/b.bin.{result.sha1[:8]}",
+        )
+        # Copied under the declared name, never over a file, and only once.
+        card = self._bios_file("card/B.BIN", missing)
+        named = validate_pr.validate_file(card, db, EMPTY_HASHES, known)
+        self.assertEqual(validate_pr.place([named]), [named.destination])
+        self.assertEqual((self.tmp / named.destination).read_bytes(), missing)
+        self.assertTrue((self.tmp / card).is_file())
+        self.assertEqual(validate_pr.place([named]), [])
+        unknown = validate_pr.validate_file(self._bios_file("card/x.bin"), db, EMPTY_HASHES, known)
+        self.assertEqual((unknown.outcome, unknown.destination), ("unknown", ""))
         placed = validate_pr.validate_file(
             self._bios_file("bios/Other/Place/b.bin", missing), db, EMPTY_HASHES, known,
         )
@@ -243,20 +259,38 @@ class HashRecognition(_Fixture):
         self.assertIn(".variants/known.bin.", markdown)
 
     def test_an_unreferenced_file_asks_for_review(self):
+        """Random bytes read "PASS" and "all checks passed", and the warning
+        named the platform configs alone though the profiles were searched."""
         rel = self._bios_file("bios/Sony/PS/mystery.bin")
-        markdown = validate_pr.validate_file(rel, None, EMPTY_HASHES).to_markdown()
+        result = validate_pr.validate_file(rel, None, EMPTY_HASHES)
-        self.assertIn("needs manual review", markdown)
+        markdown = result.to_markdown()
+        self.assertEqual(result.outcome, "unknown")
+        self.assertIn("### REVIEW", markdown)
+        self.assertIn("no platform list or emulator profile declares this file", markdown)
+        self.assertIn("which emulator reads it", markdown)
 
     def test_a_duplicate_of_a_known_file_is_reported(self):
-        rel = self._bios_file("bios/Sony/PS/dupe.bin")
+        """A copy of a held file passed, and was sent to the folder that
+        already holds it."""
+        rel = self._bios_file("elsewhere/dupe.bin")
         probe = validate_pr.validate_file(rel, None, EMPTY_HASHES)
         db = {"files": {probe.sha1: {"path": "bios/Sony/PS/original.bin"}}}
-        markdown = validate_pr.validate_file(rel, db, EMPTY_HASHES).to_markdown()
+        known = {**EMPTY_HASHES, "sha1": {probe.sha1},
-        self.assertIn("Duplicate", markdown)
+                 "folders": {probe.sha1: {"bios/Sony/PS/"}}}
-        self.assertIn("original.bin", markdown)
+        result = validate_pr.validate_file(rel, db, known)
+        markdown = result.to_markdown()
+        self.assertEqual(result.outcome, "held")
+        self.assertIn("### REVIEW", markdown)
+        self.assertIn("Already held at `bios/Sony/PS/original.bin`: nothing to add", markdown)
+        self.assertNotIn("not under bios/", markdown)
         # A duplicate is worth saying out loud but must not block the PR.
-        self.assertTrue(validate_pr.validate_file(rel, db, EMPTY_HASHES).passed)
+        self.assertTrue(result.passed)
 
+    def test_a_copy_in_a_tree_read_by_path_is_kept(self):
+        rel = self._bios_file("bios/ScummVM/a/dupe.bin")
+        probe = validate_pr.validate_file(rel, None, EMPTY_HASHES)
+        db = {"files": {probe.sha1: {"path": "bios/ScummVM/b/dupe.bin"}}}
+        self.assertNotEqual(validate_pr.validate_file(rel, db, EMPTY_HASHES).outcome, "held")
 
     def test_the_file_the_database_holds_is_not_its_own_duplicate(self):
         """The documented example on a tracked file warned "Duplicate:
@@ -334,7 +368,8 @@ class Placement(_Fixture):
 class Reporting(_Fixture):
     def test_markdown_carries_every_digest_and_the_verdict(self):
         rel = self._bios_file("bios/Sega/Saturn/sega.bin")
-        result = validate_pr.validate_file(rel, None, EMPTY_HASHES)
+        probe = validate_pr.validate_file(rel, None, EMPTY_HASHES)
+        result = validate_pr.validate_file(rel, None, {**EMPTY_HASHES, "sha1": {probe.sha1}})
         markdown = result.to_markdown()
         for value in (result.sha1, result.md5, result.crc32):
             self.assertIn(value, markdown)
@@ -344,12 +379,33 @@ class Reporting(_Fixture):
         rel = self._bios_file("bios/Sega/Saturn/bad.exe")
         self.assertIn("### FAIL", validate_pr.validate_file(rel, None, EMPTY_HASHES).to_markdown())
 
+    @unittest.skipUnless((REPO_ROOT / "database.json").is_file(), "database.json is not built")
+    def test_the_status_line_names_what_is_left_to_review(self):
+        import shutil
+        import subprocess
+
+        held = REPO_ROOT / "bios" / "Sony" / "PlayStation Portable" / "ppge_atlas.zim"
+        if not held.is_file():
+            self.skipTest(f"{held} is not here")
+        shutil.copy(held, self.tmp / "ppge_atlas.zim")
+        (self.tmp / "mystery.bin").write_bytes(os.urandom(4096))
+        proc = subprocess.run(
+            [sys.executable, str(REPO_ROOT / "scripts" / "validate_pr.py"), "--markdown",
+             str(self.tmp / "mystery.bin"), str(self.tmp / "ppge_atlas.zim")],
+            capture_output=True, text=True, timeout=300, check=False,
+        )
+        self.assertEqual(proc.returncode, 0, proc.stderr)
+        self.assertIn("**Status**: no check failed, to review: 1 already held, "
+                      "1 not recognised", proc.stdout)
+        self.assertNotIn("all checks passed", proc.stdout)
+
 
 class PlatformHashLoading(_Fixture):
     def test_missing_directory_yields_empty_sets(self):
         known = validate_pr.load_platform_hashes(str(self.tmp / "absent"))
         self.assertEqual(known, {"sha1": set(), "md5": set(), "sha256": set(), "crc32": set(),
-                                 "names": set(), "folders": {}, "sizes": {}, "entries": {}})
+                                 "names": set(), "folders": {}, "sizes": {}, "entries": {},
+                                 "named": {}})
 
     def test_declared_hashes_and_names_are_collected(self):
         platforms = self.tmp / "platforms"

tests/test_validate_pr_flow.py

@@ -13,6 +13,7 @@ import subprocess
 import sys
 import tempfile
 import unittest
+from unittest import mock
 from pathlib import Path
 
 import yaml
@@ -146,5 +147,45 @@ class RunFromAnyDirectory(unittest.TestCase):
         self.assertIn("Already in the database at this path", messages)
 
 
+class AnUnfetchedUpstreamIsFetched(unittest.TestCase):
+    """CONTRIBUTING has the project added as `upstream` with no fetch. The
+    remote named the project, so nothing was fetched, upstream/main did not
+    exist, origin/main was the pushed branch, and --changed found nothing."""
+
+    def _git(self, cwd: Path, *args: str) -> None:
+        subprocess.run(["git", *args], cwd=cwd, check=True, capture_output=True, text=True)
+
+    def test_the_branch_file_is_found(self):
+        with tempfile.TemporaryDirectory(dir=REPO_ROOT / "tmp") as tmp:
+            root = Path(tmp)
+            project = root / "Abdess" / "retrobios.git"
+            fork = root / "fork.git"
+            clone = root / "clone"
+            for bare in (project, fork):
+                bare.mkdir(parents=True)
+                self._git(bare, "init", "-q", "--bare", "-b", "main")
+            clone.mkdir()
+            self._git(clone, "init", "-q", "-b", "main")
+            self._git(clone, "-c", "user.name=t", "-c", "user.email=t@t", "commit", "-q",
+                      "--allow-empty", "-m", "base")
+            self._git(clone, "push", "-q", str(project), "main")
+            (clone / "bios").mkdir()
+            (clone / "bios" / "new.bin").write_bytes(b"x")
+            self._git(clone, "add", "bios/new.bin")
+            self._git(clone, "-c", "user.name=t", "-c", "user.email=t@t", "commit", "-q",
+                      "-m", "add")
+            self._git(clone, "remote", "add", "origin", str(fork))
+            self._git(clone, "push", "-q", "origin", "main")
+            self._git(clone, "fetch", "-q", "origin")
+            self._git(clone, "remote", "add", "upstream", str(project))
+            cwd = os.getcwd()
+            os.chdir(clone)
+            try:
+                with mock.patch.object(validate_pr, "PROJECT_URLS", (str(project),)):
+                    self.assertEqual(validate_pr.get_changed_files(), ["bios/new.bin"])
+            finally:
+                os.chdir(cwd)
+
+
 if __name__ == "__main__":
     unittest.main()

wiki/architecture.md

@@ -180,7 +180,11 @@ A platform and an emulator profile can each name a different file for the same
 place in the BIOS folder: RetroBat puts its own `ATARIOSB.ROM` at
 `bios/ATARIOSB.ROM`, where atari800 reads another dump. A claim counts only when
 it rests on content or on the path. A file found by its name alone claims
-nothing. The pack then answers to whoever asked for it:
+nothing, unless its copy has the exact size the profile declares and the
+platform's file has another size (`size_exact`): RetroDECK's `sgb2.boot.rom` is
+the Super Game Boy 2 SNES program ROM where supersnes9x reads a 256-byte boot ROM.
+Two files of equal size are revisions, not a conflict. The pack then answers to
+whoever asked for it:
 
 - A **platform pack** must leave the frontend's own check green, since a user
   reading red concludes the pack is broken. Where the frontend only looks for
@@ -280,17 +284,16 @@ being intact: see [verifying a release](release-process.md#verifying-a-release).
 
 Three things make that hold. Generated members (`README.txt`, `manifest.json`)
 carry a fixed date rather than the wall clock, and the manifest's `generated`
-field is read from `database.json` instead of the build time. MAME and FBNeo
+field is read from `database.json` instead of the build time. Each ZIP of the
-romsets are rebuilt deterministically from their ROMs, so a pack does not
+collection, MAME and FBNeo romsets included, is copied byte for byte, so it
-inherit whatever metadata the source archive happened to carry. And every
+hashes to the SHA1 the install list and the site publish, and git fixes those
-member is written with the same fixed date: `ZipFile.write` copies the source
+bytes. And every member is written with the same fixed date: `ZipFile.write`
-file's mtime, which is the checkout time for a file from the collection and
+copies the source file's mtime, which is the checkout time for a file from the
-the wall clock for one the build just staged in `tmp/`.
+collection and the wall clock for one the build just staged in `tmp/`.
 
 `tests/test_deterministic_zip.py` builds the same fixture twice and compares
 the bytes, for both the platform and the emulator pack paths. Its fixture
-holds a romset, because without one the comparison never reaches the rebuild
+holds a ZIP, so the comparison covers the archive copy as well as loose files.
-path and passes while the real packs still move.
 
 ## Collection rules
 
@@ -383,7 +386,8 @@ resolution.
   deploy identity. No workflow builds or publishes a release.
 
 - `safe_extract_zip()` prevents zip-slip path traversal attacks
-- `deterministic_zip` rebuilds MAME ZIPs so same ROMs always produce the same hash
+- a pack copies each ZIP of the collection byte for byte, so every member
+  hashes to the SHA1 the install list publishes
 - `crypto_verify.py` and `sect233r1.py` verify 3DS RSA-2048 signatures and AES-128-CBC integrity
 - ZIP inner ROM verification via `checkInsideZip()` replicates Batocera's behavior
 - `md5_composite()` replicates Recalbox's composite ZIP hash

wiki/reading-reports.md

@@ -140,8 +140,13 @@ are still included in packs automatically. A required one the collection lacks
 is listed under the counts, and `--verbose` lists the optional ones too.
 
 The cross-reference uses `resolve_platform_cores()` to determine which emulator
-profiles are relevant for each platform, then checks whether each profile's files
+profiles are relevant for each platform, then checks whether the platform
-appear in the platform config.
+already declares each profile's files (`declaration_settles()`). A name alone
+is not enough. The declaration must fill the core's own slot or have a size the
+core accepts: Galaksija's 4 KiB `ROM1.BIN` does not stand for DOSBox's 32 KiB
+SC-55 one. An alias counts only at the path where the core reads it, so
+RetroDECK's root `MSX.ROM` is not CLK's `MSX/MSX.ROM`. An archive counts by its
+name.
 
 ### Filtering output
 

wiki/release-process.md

@@ -166,9 +166,9 @@ PR based on changed paths:
 ## Large files management
 
 Files too large for git are stored as assets on a permanent GitHub release
-named `large-files`. Size does not decide it: four committed files are over
+named `large-files`. Size does not decide it: four committed files are
-50 MB (`history.xml`, `default-GM.sf3`, `5.avi`, `tbblue.mmc`) and are served
+over 50 MB (`history.xml`, `default-GM.sf3`, `5.avi`, `uqm-remix-disc1.uqm`)
-from the repository like any other.
+and are served from the repository like any other.
 
 Examples: PS3UPDAT.PUP, PSVUPDAT.PUP, PSP2UPDAT.PUP, the DSi NAND images,
 maclc3.zip, Firmware.19.0.0.zip (Switch), the QEMU EDK2 firmware, the ScummVM

wiki/tools.md

@@ -15,7 +15,7 @@ Run everything in sequence:
 
 ```bash
 python scripts/pipeline.py --offline              # DB + verify + packs + manifests + integrity + readme + site
-python scripts/pipeline.py --offline --skip-packs  # DB + verify + readme + site
+python scripts/pipeline.py --offline --skip-packs  # no packs or manifests
 python scripts/pipeline.py --offline --skip-docs   # skip readme + site generation
 python scripts/pipeline.py --offline --target switch  # filter by hardware target
 python scripts/pipeline.py --offline --with-truth  # include truth generation + diff
@@ -63,8 +63,8 @@ Pipeline flags:
 | Flag | Effect |
 |------|--------|
 | `--offline` | skip every step that needs the network (2, 2a, 2a2, 2b, 2b2), and pack only the release assets already in `.cache/large/` |
-| `--skip-packs` | skip steps 4, 4b, 4c, 6 and the consistency check. README and site are still rewritten unless `--skip-docs` is given too |
+| `--skip-packs` | skip steps 4, 4b, 4c, 5 and 6. Steps 4d and 4e still run, and README and site are still rewritten unless `--skip-docs` is given too |
-| `--skip-docs` | skip steps 7 and 8 |
+| `--skip-docs` | skip steps 7, 8 and 8b |
 | `--include-archived` | include archived platforms in verify, packs, truth and export |
 | `--target TARGET` | filter verify, packs and truth by hardware target |
 | `--source {platform,truth,full}` | pack file source, passed through to `generate_pack.py`. Refused with `--skip-packs` |
@@ -1334,7 +1334,7 @@ before a release (wiki/release-process.md). Its lines stay in
 | `validate_site.py` | Validate rendered metadata, heading order, image alternatives, JSON-LD, local resources, links and fragments, and the page size and search index budgets |
 | `romset_recipes.py` | Identify which emulator version an arcade archive matches, and rebuild a pinned archive from ROMs already held |
 | `scraper/romset_dat_importer.py` | Fetch and import per-set recipes from MAME `-listxml` or FBNeo DATs into `recipes/`. `--fetch` alone takes the newest MAME release, and FBNeo DATs are refreshed by git blob sha. `--fetch <revision>` reads FBNeo at that tag, branch or commit |
-| `deterministic_zip.py` | Rebuild MAME BIOS ZIPs deterministically (same ROMs = same hash) |
+| `deterministic_zip.py` | The fixed date pack members carry. Its archive rebuild is no longer called by the pack, which copies each ZIP byte for byte |
 | `torrentzip.py` | Build TorrentZip archives for MAME/FBNeo ROM sets (archive bytes depend only on contents) |
 | `crypto_verify.py` | 3DS RSA signature and AES crypto verification |
 | `sect233r1.py` | Pure Python ECDSA verification on sect233r1 curve (3DS OTP cert) |
@@ -1373,7 +1373,7 @@ Further options, by script:
 | `romset_recipes.py` | `--dat-recipes PATH`, a snapshot or a directory of them (default `recipes/`) |
 | `split_pack.py` | `--max-size 1900M`, the ceiling per part (K, M, G, default 2G) |
 | `torrentzip.py` | `--check ZIP` reports conformance, `--rebuild ZIP -o OUT` normalizes an archive |
-| `validate_pr.py` | `--changed` takes the BIOS files the branch changes, a folder named checks every file under it, `--markdown`, `--json` |
+| `validate_pr.py` | `--changed` takes the BIOS files the branch changes, a folder named checks every file under it, `--place` copies each recognised file from outside `bios/` to its folder and declared name (or its `.variants/` slot), `--markdown`, `--json` |
 | `validate_site.py` | `--site-dir` (rendered site, default `website/dist`), `--base` (path the site is served under), `--max-errors N` |
 | `scraper/coreinfo_scraper.py` | `--compare-db database.json` lists the `.info` firmware the database lacks |
 | `scraper/mame_hash_scraper.py` | `--release mameNNNN --profile FILE --source-commit SHA --core NAME --built-by PLATFORM` writes the profile of a MAME build a platform compiles. `--devices --listxml FILE` documents in `mame.yml` every device set with dumped ROMs that the release's `-listxml` names (`isdevice="yes"`), with `--dry-run` to preview |