A source_ref part that names a compiled image now has to carry an
address, a decompiled function or an IL method. An import directory, a
string table or readelf output only says the binary exists. A name
under a directory is a file of the tree and a .dll.map is not a DLL.
The exception list is empty: every closed profile now cites code.
validate_schemas now refuses a files[] source_ref whose parts are all data or build files (a table, a makefile, meson.build, the file itself). A ref passes when one part names a source file or a binary with an address. The closed-source profiles whose refs are still prose about the shipped binary are listed in DATA_ONLY_REF_EXCEPTIONS with a reason each.
AdvanceMAME reads its sets from dir_rom, a list of folders
(advance/osd/fileio.c:74, doc/advmame.d:680,699-701). EmuELEC's
advmame.rc sets dir_rom /storage/roms/mame (advmame.rc:11), and
advmame.sh rewrites /roms/mame to /roms/arcade when it is started for
the arcade system (advmame.sh:26-30). The extractor kept the keys under
the BIOS root alone: it now records those under the roms root too, and
a key the script rewrites gives one record for arcade and one for every
other system. A configgen record carries the systems it is written for.
The profile lays its 29 sets out at the top of dir_rom. On EmuELEC they
land in roms/mame. The 12 root archives AdvanceMAME alone claimed there
now hold the copies of the cores that also declare them (mame, geolith),
none of them lost.
A destination outside the BIOS tree is no proof of content: the moved
entry resolves by its layout path alone. roms/mame/konamigv.zip took
bios/Arcade/MAME/konamigv.zip, MAME 0.289's set, by that tail.
GZDoom collects its IWAD candidates from [IWADSearch.Directories], reads
the support WAD there and the Load files from the IWAD's own folder
(src/d_iwad.cpp:456-476, 593-624, 862-901). RetroDECK lists
RETRODECKROMSDIR/doom there (gzdoom/rd_assets/rd_config/gzdoom.ini:9),
which component_prepare.sh:27 rewrites to $roms_path/doom. The extractor
read only the BIOS placeholder of a preset: it now records every line
naming the BIOS or the roms folder the script rewrites, with that sed
line in its source. $roms_path, whose default is the roms root
(config/retrodeck/retrodeck.json:5), is read as that root.
The profile names IWADSearch.Directories as a pool and lays its IWADs
and their support and Load files out at its top. On RetroDECK the 20
held IWADs land in roms/doom, the 23 others are named missing there.
The root copies stay for prboom and uzdoom, which claim them. No other
platform hands GZDoom that list, so none changes.
MAME splits rompath, samplepath, hashpath and pluginspath on semicolons
and tries each directory in turn (src/emu/fileio.cpp:53-58, 330-392,
src/emu/emuopts.cpp:44-46, 53). The launcher records spell those lists
with the game's own folder in them, which handed_directories skipped:
- RetroBat passes -rp bios;<rom directory> (Mame64.Generator.cs:93-96)
and -rompath bios/psxmame/bios;<rom directory> to pSxMAME
(PSXMame.Generator.cs:38-48).
- KNULLI passes -rompath <romDirname>;/userdata/bios/mame;/userdata/bios
(mameGenerator.py:116-122).
- RetroPie writes rompath $romdir/mame;$romdir/arcade;$biosdir/mame
(mame.sh:133).
A semicolon list now gives each directory but the game's folder. A pool
that already lists the folder the files take keeps them there, and a
pool standing for the whole layout holds its files, never its folders.
The MAME profile names its sets at the top of the rompath and its
search settings as pools, pSxMAME its sets and rompath.
RetroBat's -sp and -hash records split on `bool hbmame = emulator ==
"hbmame"` (Mame64.Generator.cs:30): bios/mame/samples for mame64,
bios/hbmame/samples for hbmame.
Manifests: RetroBat gains the 12 pSxMAME sets in psxmame/bios and drops
samples/bbc.zip, read by nobody, which mame/samples/bbc.zip already
serves. KNULLI drops the same copy. RetroPie gains the 1427 MAME sets
in BIOS/mame, the root copies staying for the libretro cores. EmuDeck
names samples/bbc.zip as read outside the BIOS folder. RetroDECK's
samplepath is $bios_path/mame-sa/samples (mame/component_prepare.sh:94,
117): bbc.zip moves there from bios/samples/. Its MAME runs with the
-rompath ES-DE passes, %GAMEDIR%;%ROMPATH%/<system>, so the standalone
sets are named as read outside the BIOS folder and the root copies stay
for the libretro core. The 1427 sets
of RetroBat, KNULLI and EmuDeck stay at the BIOS root, listed in their
rompath. No file of any manifest is lost.
validation: [romdir] walks a PS2 ROM directory the way the profile's
fork reads it, masked_sha1 hashes a file with a byte range cleared, and
size_multiple bounds a size to a granule. The resolver name steps, the
pack and manifest scans, the emulator expansion and verify call the same
content_fits. A PCSX2 companion is taken only beside an image that
passes, and a companion placeholder no longer seeds a scan of its own.
Dolphin reads its User folder at $XDG_DATA_HOME/dolphin-emu, else
~/.local/share/dolphin-emu (Source/Core/UICommon/UICommon.cpp:449-477).
The profile names both, so a launcher handing that folder places the
layout:
- EmuELEC starts Dolphin with XDG_DATA_HOME=/storage/roms/dolphin
(dolphinSA/scripts/dolphin.sh:28): its 30 files move from the BIOS
root to dolphin/dolphin-emu/ of the install tree.
- RetroDECK links User/Triforce to bios/Triforce
(dolphin/component_prepare.sh:30, 49), where its list already puts
segaboot.gcm. The rest of the User folder sits in the Flatpak's data
directory and is named as read outside the BIOS folder.
An entry only the standalone build reads, path and no standalone_path,
now takes that path as its layout path (Load/Firmware/rtl_bt/), and
the pass copying a name the platform declares elsewhere reaches an
unprefixed path: gba_bios.bin lands in dolphin/dolphin-emu/GBA/.
RetroDECK sets mlc_path to $bios_path/cemu (cemu/component_prepare.sh:14,
29), which replaces the MLC Cemu reads under its user data directory
(src/config/CemuConfig.cpp:25-26, ActiveSettings.cpp:295-320). The
profile names that key, so the layout's mlc01/ folder is placed in
bios/cemu. Two readers kept the binding from reaching the manifest:
- a moved entry kept its standalone load_from, data_dir, and the gap
report set the MLC aside as read outside the system directory. The
platform's folder now replaces the build's own root.
- an entry read outside the BIOS tree was no longer offered to the
launcher layer, so RetroDECK's link of keys.txt to bios/cemu/keys.txt
went unserved. It is offered again, fills no destination by its own
path, and a launcher place it fills settles it.
The RetroDECK manifest gains the 55 MLC files under bios/cemu/. It
drops the root copies of keys.txt, otp.bin, seeprom.bin,
shareddata.bin, cafeLibs/ and resources/sharedFonts/, which standalone
Cemu reads in the Flatpak's $XDG_DATA_HOME/Cemu and names them as read
outside the BIOS folder. The libretro copies under bios/Cemu/ stay.
EmuELEC links /storage/.local/share/flycast to /storage/roms/bios/dc
(flycast.sh:11-14 at v4.8), and Flycast reads
$HOME/.local/share/flycast. Root's home there is /storage
(packages/sysutils/busybox/package.mk:168, profile.d/98-busybox.conf:4,
sourced by flycast.sh:7), which nothing recorded, so the link bound
nothing and flycastsa's files went to the BIOS root.
The registry's home_directory now spells that directory as ~ in the
handles and directories of launcher records. flycastsa's ten files
land in bios/dc, where the libretro build already reads them. The
manifest drops dc_boot.bin and dc_nvmem.bin from the BIOS root, which
no emulator reads there, and six arcade archives at the root go back
to the MAME sets the mame cores claim.
Two filename-agnostic cores reading one BIOS folder each scan the same
images. The second scan stopped at an image the first had placed, so a
companion only the second core opens never reached the pack: Recalbox's
libretro PCSX2 reads <bios>.erom beside the images LRPS2 placed in ps2/.
The notes open on the command and the handheld app, then the full packs
for a device without internet, the setups the command alone serves, what
changed, everyone who helped since the start (read with gh, --thanks FILE
for a checked list) in a folded block, and the full changelog from the
previous release's record commit.
check_release reports a launcher in the checksum list as unrecorded and
one attached as extra, as it does for install.py. The release steps no
longer run build_port, and the checksum and upload globs name the packs
(*_BIOS_Pack*.zip) so a launcher left in dist/ is never caught.
build_port's four archives carry install.py pinned by SHA-256, so they
change with the installer, not with the packs. The site now builds them
from its own commit and serves them under get/handheld/. Each get/ page
of Batocera, KNULLI, ROCKNIX and AmberELEC names its file, size, hash,
folder and menu, and get/linux-handhelds/ does the same for muOS, ArkOS,
dArkOS and JELOS. The installer's handheld section opens on a download
table, both tables written from sitehandheld.LAUNCHERS.
KNULLI sets AetherSX2's [Folders] Bios to bios/ps2
(aethersx2Generator.py:219), yet its 112 PS2 BIOS files went to the
BIOS root. Three causes:
- a record's passes_to was split into words, so "[Folders] Bios"
could not be told from "[Filenames] BIOS". ini_keys() adds the
section-qualified key a profile names in standalone_directories.
- scans_dir records hand a folder too (ROCKNIX's bios/armsx2,
EmuDeck's PCSX2 and ARMSX2) and were not read as handed.
- the agnostic scan looked its seeds up by the id of the profile's
entries, while it walks the copies platform_layouts makes: any
bound layout scanned nothing. Copies map back to the profile's
entries by position.
An agnostic profile's entry the layout reads outside the BIOS tree
is now named by the gap report (outside_bios_folder) instead of
dropped: ARMSX2's arcade/proverb.zip on ROCKNIX and EmuDeck.
Batocera 43.1 builds libretro/ps2 at 416291a with 001.bios-dir.patch and
Recalbox 10.1.1 builds it at 9485a53 with 0001-recalbox-paths.patch. Both
patches move the BIOS scan from pcsx2/bios to ps2, and Recalbox also moves
the data root to system/ps2. Each build gets its own built_by profile:
lrps2-416291a under Batocera's pcsx2 name, lrps2-9485a53 under Recalbox's
ps2 name.
A platform's own build now answers for its own build mode only on a
dual_mode_cores name, so Batocera keeps standalone PCSX2 beside its
libretro build.
A profile standalone copy was dropped when the manifest held its file
under two destinations. With the libretro PS2 cores' pcsx2/bios/ scan,
DuckStation's ps2_scph50006.bin and ps2_scph50008.bin copies left the
Batocera, Recalbox and RetroBat manifests. Same bytes at every path
are one source: the copy whose cores name the emulator, else the
shortest path.
LRPS2 and PCEE2 name their BIOS <bios>.bin, which resolves to no
file, so the platform pack's agnostic scan never anchored pcsx2/bios/
for them. On RetroArch yaps2 anchored it and the files carried only
yaps2 as their core. On Batocera and RetroBat, where yaps2 does not
run, pcsx2/bios/ stayed empty or held only the platform's own list.
The scan now takes the folders where the system's other profiles
resolve a hashed dump of that shape, the evidence the emulator pack
already used, and a companion (<bios>.nvm) looks where its image is
proven.
The release carries the packs, the handheld launchers and the signed checksum list. The installer and the file lists are pinned from the tag (installer page, For packagers), and check_release.py now refuses them as unrecorded files. The notes link each pack's files on the release itself, in a beginner's page: one idea per line. The handheld section links the latest release.
An entry the installer places in the roms, emulators or configs folder is in the install list but not in the ZIP, which is unzipped into the BIOS folder. pack_files and pack_size counted them: release_record.py refused Batocera (19196 against 19125 in the archive) and RetroBat (9200 against 8882). Manifests regenerated twice, identical.
A query word matches the start of a longer word. Since the index grew,
the "2" of "playstation 2 bios" matched the 2026 of every page's date
line and the "u" of "wii u bios" the "use" of Next steps, so the
PlayStation and Nintendo pages held every query word and ranked above
the PlayStation 2 and Wii U pages. Both blocks carry
data-pagefind-ignore now: search_gate answers 58 of 58 queries, the
four console pairs by 9 to 14 percent. Batocera's pages weigh 882
bytes more for it.
KNULLI builds amiberry from Batocera's package through its batocera
submodule: the reference now names knulli-linux at 20260511, the tag
platforms/knulli.yml is read at, and the package file below the
submodule. check_freshness follows a path .gitmodules declares to the
commit the tree pins, which reads AMIBERRY_VERSION = v7.1.0 there.
RetroBat pins amiberry 8.3.0.0 only in emulators/versions.xml on its
download server, read by emulatorLauncher/Installer.cs. No repository
holds it, so amiberry-8.3.0 keeps pin_frozen, which says so, and no
pin_build.
The PlayStation section opened on psxonpsp660.bin under 'every
region': since sets carry the requirement, ARMSX1's region-free member
of its required BIOS set counts as required, and the world line came
first, above the scph files most emulators read, and its card first.
A file read for every region is a stand-in for each region's BIOS, so
that line now follows the territories.
The Apple II start now leads on BizHawk's and Mednafen's files, a tie
of two readers with CLK's and POM2's apple2e.rom. The test named that
one file. It now checks that every start file is a link, its intent.
A release asset absent from the checkout is omitted from the install
list. That omission named only the first core reading it: harm1.wad,
action2.wad and delaweare.wad named gzdoom, so install.py --target
rpi1, where lzdoom runs, kept nothing the builder's list for rpi1
omits. The omission now names the cores of the claims folded onto the
entry too.
Every error outside mypy-baseline.txt is fixed in the code:
- common.py: the CRC32 step reused found, the name the layout branch
gives a resolution tuple, for its list of matches. It now has its own
name. bindings_of annotates its systems.
- launchers.py, retropie_launchers.py, install.py: an Optional value
read before its None check.
- recalbox_launchers.py, rocknix_launchers.py: Launcher calls take
typed keywords, a partial or named parameters, instead of a
dict[str, object] spread. end_lineno falls back to lineno.
- retrobat_launchers.py: the StageRomKey match no longer reuses the
str key, and record values are checked before they are iterated.
- batocera_launchers.py: the value of an assignment is read on an
Assign or AnnAssign only, and the launcher list is annotated.
- emuelec_launchers.py, emuelec_scraper.py, rocknix_scraper.py: import
fallbacks no importer reaches are dropped, a reused name renamed.
- batocera_targets_scraper.py: a submodule answer without a commit is
refused.
The baseline loses the 30 entries these fixes resolved and gains
none.
An executable is refused unless an emulator profile pins its exact bytes by sha1 or sha256. PICO-8's pico8.exe and SDL2.dll, from the official 0.2.7 archive, are pinned by the lexaloffle profile; md5 and crc32 never vouch for code.
KNULLI's system pages linked /emulators/test/ from the cores examined for an entry: the test core resolves as a downloadable core and has no page, which failed the site's link check. Cores examined now link only the profiles the site writes a page for.
The xemu generator writes eeprom.bin and xbox_hdd.qcow2 into
saves/<system> from its own resources and passes both in xemu.toml,
copies chihiro_eeprom.bin between saves/chihiro and the emulator folder,
and points the Chihiro snapshot store at saves/chihiro. The Cxbx
generator runs each build from its emulator folder (chihiro, chihiro-ds,
cxbx-reloaded, else cxbx-r), where it writes settings.ini, writes
EEPROM.bin into cxbx-reloaded for an Xbox, and mounts an .iso with its
own xbox-iso-vfs.exe after checking the Dokan 2 runtime. Nine records,
read at emulatorLauncher ca486005.
A program the launcher runs from its own folder is no launcher read
left to collect. xemu-chihiro lists the two files it generates,
chihiro_eeprom.bin and chihiro_snapshots.qcow2, as unsourceable.
The amiberry and hatarib extractors give a search record and its copy
record one list object, and PyYAML wrote it as &id001 and *id001.
platforms/retrobat.yml, batocera.yml and knulli.yml carried such
anchors. The platform dumper now writes every list out, and the two
files the RetroBat scrape does not rewrite are dumped again through it:
their content is unchanged.
RetroPie lists KS 1.3 as kick34005.A500, and its fs-uae launcher
passes $biosdir/kick13.rom. The listed copy settles fs-uae's entry by
content, so the launcher layer never offered it, and kick13.rom
reached the pack only as a copy of UAE4All2's amiga/kick13.rom: the
x86 boards, without UAE4All2, lost it. An entry the list settles under
another name now answers the launcher's place by its own name.
The first profile reading a requirement hosts the others' claims,
and two ways of dropping the host dropped them too. DraStic brings its
own usrcheat.dat, and DSPerate, which reads the same file, lost it
wherever DraStic runs. A missing file was recorded for its first
reader alone: NAPALM.GRP named Raze, Data.rsdk Sonic Mania. A list
built for a board without that reader carried what the full list
left out, and install.py --target disagreed with the builder on
ROCKNIX, KNULLI, Batocera and RetroPie boards.
A board's core list names Batocera's standalone amiberry and the
buildbot's amiberry core alike, so install.py --target kept the
amiberry 8.5.0 libretro files on bcm2711, whose RetroArch downloads
nothing. resolve_platform_cores decides that reach, and the install
list now carries its answer as downloaded_cores, read by the
installer's narrowing for a target.
The builder tested only the first installable name of a profile:
Batocera lists higan_sfc_balanced and nside_sfc_balanced, the x86_64
downloader offers the second, and the core left the x86_64 pack with
its Super GameBoy files.
RetroPie builds amiberry 5.7.2 and 6.3.3, one per board, both in
$md_inst of its amiberry package. Each build recorded roms/kick.rom
and whdboot/AmiQuit in the emulator's folder apart, and
install/retropie.json listed both destinations twice. A file read
outside the BIOS tree is now keyed by the folder the platform installs
the listed name into, and one omission names every core reading it.
Since ddafc2ff4 a pack's manifest.json holds install_list and
targets, which install.py reads with the pack alone. The pack schema
refused both. It now names them, and validate_schemas.py checks each
against the install and target manifest schemas.
The ZIP replay in packextras.hidden_extras now skips roms, emulators and configs entries as the builder does, so verify and pack integrity stop counting files only the installer places. The consistency check treats the Android .nomedia marker as a pack note. MAME4droid's hiscore.dat leaves bios/: the app unpacks it from its own files.zip. VVVVVV's Make and Play data.zip is collected as a release asset, which EmuELEC declares as required.
The plan, Confirm and the result say what happens, its size, where and
the next step. Their counts, free space, backup and save notes, kinds,
reasons a file is left out, per-file failures and commands move to a
Details screen: d on the plan, Confirm and the result, the Details row
on the plan, Y, o or Tab on the result. Opened on the result, they are
also said in the terminal after Exit. Installing lists the files that
went well only with --verbose, and --verbose draws every line in place.
The log line reads Log: so that Details names one thing.
--check prints its count, the folder and the remedy in both modes, so
that --verbose prints every default line. Each golden screen gets a word
budget per size, and a test proves the screen and its Details together
equal the --verbose screen.
By default a run prints what it installs, the BIOS folder, the progress
at a terminal, the result and the next step. The steps, the per-file log
lines, the counts and the reasons a file is left out come back with
--verbose. --check gives the count and the folder, and lists each file
with --verbose. The installer page shows the default run and walks
through the --verbose one.
No platform or profile files a system under neo-geo-pocket-color since profiles name it snk-ngpc: its SYSTEM_PARENTS entry pointed at an id without a name.
bundled_only.py --strict listed 14 copies in bios/Commodore/Amiga/
.variants that only the per-build amiberry profiles reach, each through
a bundled entry. Four are the AROS ROMs, and those entries say
not_bundled_on: ROCKNIX, AmberELEC and Recalbox package amiberry
without them, so their packs bring them from the collection. The tool
read bundled alone and listed them, and now keeps a copy a
not_bundled_on entry reaches. The ten others (WHDLoad, boot-data.zip,
whdload_db.xml of 5.6.5 to 8.0.0) no platform, profile or manifest
needs: removed. database.json is to be regenerated with the
collection's release assets.
Since the groups model, load_emulator_profiles marks every member of a
required set required, and the page counted each one: DuckStation,
which needs one BIOS of the game's region among the 105 images of its
ps-bios set, read "105 required, 0 optional". The site counts a
required set once and says one file fills it, and a set whose members
the code picks by region reads as one choice. The tests were not keyed
by a count but by the profile's name, through raw YAML that no longer
carries the derived requirement: they now read the profile as the site
does, take the count from the set, and keep their guard against a
get-all-of-them reading. Fourteen other profiles with required sets
(kronos, pom2, picodrive...) were overcounted the same way.
Abdessamad Derraz