Successful Windows/MSVC and Ubuntu/Ninja CI jobs publish these workflow artifacts:
marc-interoperability-windows-msvc-x64
marc-interoperability-ubuntu-ninja-x64
Each current schema-58 bundle contains the same generated input.bin, the
frozen 42 stable-profile archives, twenty-five experimental Format 2 archives,
one position-distance archive, and manifest.json. The manifest declares
codec set marc-cli-v58 and records
the source revision, producing platform, compiler label, architecture, CLI
SHA-256, and the size and SHA-256 of every input and archive file.
Download and extract a bundle from a successful GitHub Actions run. Build marc at the same commit on the platform being tested, then use an output directory that does not already exist:
powershell.exe -NoProfile -ExecutionPolicy Bypass -File `
tests/verify_interoperability_bundle.ps1 `
-MarcCli build-msbuild/Release/marc.exe `
-BundleDirectory downloaded-bundle `
-OutputDirectory out/interop-checkOn a host with PowerShell 7, use pwsh -NoProfile -File with the same
arguments. The verifier performs all of the following:
- validates the manifest version, exact codec set and profile order, leaf-only file names, sizes, and SHA-256 values;
- decodes all sixty-eight foreign archives and compares their output byte
for byte with
input.bin; - re-encodes
input.binwith the local executable and compares every complete archive byte for byte with the foreign archive.
Report the producing artifact name, local OS and architecture, local compiler, tested commit, final verifier line, and any failure output. A successful report has this form:
artifact: marc-interoperability-windows-msvc-x64
local platform: <OS, architecture, compiler>
commit: <manifest source_revision and local Git commit>
result: Verified 68 archives from windows-msvc-x64 (...), revision <Git object ID>
The verifier remains able to validate legacy schema-1 bundles with their exact
seven-profile set, schema-2 bundles with marc-cli-v2 and exactly eight
archives, schema-3 bundles with marc-cli-v3 and exactly thirteen archives,
and schema-4 bundles with marc-cli-v4 and exactly fifteen archives. Schema 5
requires marc-cli-v5 and all sixteen archives, appending
lzw-blocked-huffman to the frozen schema-4 order. Schema 6 requires
marc-cli-v6 and all seventeen archives, appending lzd-blocked-huffman to
the frozen schema-5 order. Schema 7 requires marc-cli-v7 and all eighteen
archives, appending lzmw-blocked-huffman to the frozen schema-6 order. Schema
8 requires marc-cli-v8 and all nineteen archives, appending
lz77-adaptive-huffman to the frozen schema-7 order. Schema 9 requires
marc-cli-v9 and all twenty archives, appending lzss-adaptive-huffman to the
frozen schema-8 order. Schema 10 requires marc-cli-v10 and all twenty-one
archives, appending lz78-adaptive-huffman to the frozen schema-9 order.
Schema 11 requires marc-cli-v11 and all twenty-two archives, appending
lzw-adaptive-huffman to the frozen schema-10 order. Schema 12 requires
marc-cli-v12 and all twenty-three archives, appending
lzd-adaptive-huffman to the frozen schema-11 order. Schema 13 requires
marc-cli-v13 and all twenty-four archives, appending
lzmw-adaptive-huffman to the frozen schema-12 order. Schema 14 requires
marc-cli-v14 and all twenty-five archives, appending lz77-dynamic-range to
the frozen schema-13 order. Schema 15 requires marc-cli-v15 and all
twenty-six archives, appending lzss-dynamic-range to the frozen schema-14
order. Schema 16 requires marc-cli-v16 and all twenty-seven archives,
appending lz78-dynamic-range to the frozen schema-15 order. Schema 17
requires marc-cli-v17 and all twenty-eight archives, appending
lzw-dynamic-range to the frozen schema-16 order. Schema 18 requires
marc-cli-v18 and all twenty-nine archives, appending lzd-dynamic-range to
the frozen schema-17 order. Schema 19 requires marc-cli-v19 and all thirty
archives, appending lzmw-dynamic-range to the frozen schema-18 order. Schema
20 requires marc-cli-v20 and all thirty-one archives, appending lz77-rans
to the frozen schema-19 order. Schema 21 requires marc-cli-v21 and all
thirty-two archives, appending lzss-rans to the frozen schema-20 order.
Schema 22 requires marc-cli-v22 and all thirty-three archives, appending
lz78-rans to the frozen schema-21 order. Schema 23 requires marc-cli-v23
and all thirty-four archives, appending lzw-rans to the frozen schema-22
order. Schema 24 requires marc-cli-v24 and all thirty-five archives,
appending lzd-rans to the frozen schema-23 order. Schema 25 requires
marc-cli-v25 and all thirty-six archives, appending lzmw-rans to the frozen
schema-24 order. Schema 26 requires marc-cli-v26 and all thirty-seven
archives, appending lz77-tans to the frozen schema-25 order. Schema 27
requires marc-cli-v27 and all thirty-eight archives, appending lzss-tans
to the frozen schema-26 order. Schema 28 requires marc-cli-v28 and all
thirty-nine archives, appending lz78-tans to the frozen schema-27 order.
Schema 29 requires marc-cli-v29 and all forty archives, appending lzw-tans
to the frozen schema-28 order. Schema 30 requires marc-cli-v30 and all
forty-one archives, appending lzd-tans to the frozen schema-29 order. Schema
31 requires marc-cli-v31 and all forty-two archives, appending
lzmw-tans to the frozen schema-30 order. Schema 32 requires marc-cli-v32
and all forty-three archives, appending the experimental
lzss-contextual-dynamic-range archive to the frozen schema-31 order. Schema
33 requires
marc-cli-v33 and all forty-four archives, appending the experimental
lzss-contextual-rans-compact archive to the frozen schema-32 order; the
fixed-descriptor lzss-contextual-rans diagnostic remains absent. Schema 34
requires marc-cli-v34 and all forty-five archives, appending the experimental
lzss-contextual-tans archive to the frozen schema-33 order. Schema 35 requires
marc-cli-v35 and all forty-six archives, appending the experimental
lzss-contextual-blocked-huffman archive to the frozen schema-34 order. Schema
36 requires marc-cli-v36 and all forty-seven archives, appending the
experimental lzss-contextual-adaptive-huffman archive to the frozen
schema-35 order. Schema 37 requires marc-cli-v37 and the same forty-seven
archive bytes and order, but renames archive 44's manifest codec and leaf from
the historical lzss-contextual-rans-compact to the canonical
lzss-contextual-rans. Schema 38 requires marc-cli-v38 and all forty-eight
archives, appending lzss-contextual-dynamic-range-1m to the frozen schema-37
order. Schema 39 requires marc-cli-v39 and all forty-nine archives, appending
lzss-contextual-rans-1m to the frozen schema-38 order. Schema 40 requires
marc-cli-v40 and all fifty archives, appending
lzss-contextual-tans-1m to the frozen schema-39 order. Schema 41 requires
marc-cli-v41 and all fifty-one archives, appending
lzss-contextual-blocked-huffman-1m to the frozen schema-40 order. Schema 42
requires marc-cli-v42 and all fifty-two archives, appending
lzss-contextual-adaptive-huffman-1m to the frozen schema-41 order. Schema 43
requires marc-cli-v43 and all fifty-three archives, appending
lzss-contextual-dynamic-range-4m to the frozen schema-42 order. Schema 44
requires marc-cli-v44 and all fifty-four archives, appending
lzss-contextual-rans-4m to the frozen schema-43 order. Schema 45 requires
marc-cli-v45 and all fifty-five archives, appending
lzss-contextual-tans-4m to the frozen schema-44 order. Schema 46 requires
marc-cli-v46 and all fifty-six archives, appending
lzss-contextual-blocked-huffman-4m to the frozen schema-45 order. Schema 47
requires marc-cli-v47 and all fifty-seven archives, appending
lzss-contextual-adaptive-huffman-4m to the frozen schema-46 order. No schema
silently inherits profiles or names added by a later schema.
Schema 48 requires marc-cli-v48 and all fifty-eight archives, appending
lzss-contextual-dynamic-range-16m to the frozen schema-47 order. No earlier
schema silently inherits this later profile or name.
Schema 49 requires marc-cli-v49 and all fifty-nine archives, appending
lzss-contextual-rans-16m to the frozen schema-48 order. No earlier schema
silently inherits this later profile or name.
Schema 50 requires marc-cli-v50 and all sixty archives, appending
lzss-contextual-tans-16m to the frozen schema-49 order. No earlier schema
silently inherits this later profile or name.
Schema 51 requires marc-cli-v51 and all sixty-one archives, appending
lzss-contextual-blocked-huffman-16m to the frozen schema-50 order. No earlier
schema silently inherits this later profile or name.
Schema 52 requires marc-cli-v52 and all sixty-two archives, appending
lzss-contextual-adaptive-huffman-16m to the frozen schema-51 order. No
earlier schema silently inherits this later profile or name.
Schema 53 requires marc-cli-v53 and all sixty-three archives, appending
lzss-contextual-dynamic-range-64m to the frozen schema-52 order. No earlier
schema silently inherits this later profile or name.
Schema 54 requires marc-cli-v54 and all sixty-four archives, appending
lzss-contextual-rans-64m to the frozen schema-53 order. No earlier schema
silently inherits this later profile or name.
Schema 55 requires marc-cli-v55 and all sixty-five archives, appending
lzss-contextual-tans-64m to the frozen schema-54 order. No earlier schema
silently inherits this later profile or name.
Schema 56 requires marc-cli-v56 and all sixty-six archives, appending
lzss-contextual-blocked-huffman-64m to the frozen schema-55 order. No earlier
schema silently inherits this later profile or name.
Schema 57 requires marc-cli-v57 and all sixty-seven archives, appending
lzss-contextual-adaptive-huffman-64m to the frozen schema-56 order. No
earlier schema silently inherits this later profile or name.
Schema 58 requires marc-cli-v58 and all sixty-eight archives, appending
lzss-position-distance-dynamic-range as archive 68 after the unchanged
schema-57 prefix. The generated fixture and all earlier codec configurations
remain unchanged. The new archive must carry exactly format 2.0 dictionary
2/8, context 1/9, entropy 3/2. Schemas 1 through 57 keep their original sets.
On 2026-09-27 the maintainer reported successful CI and all four schema-58
external verification directions at revision
559c16a8280687f1ad57602ae01e133dbece3611, with 68 archives in every result:
| Producer | Verification host | Result |
|---|---|---|
| Windows/MSVC x64 CI | Ubuntu 26.04 / Clang 21.1.8 x64 | 68 verified |
| Ubuntu 24.04 / Ninja x64 CI | Ubuntu 26.04 / Clang 21.1.8 x64 | 68 verified |
| Ubuntu 26.04 / Clang 21.1.8 x64 | Ubuntu 26.04 / Clang 21.1.8 x64 | 68 verified |
| Ubuntu 26.04 / Clang 21.1.8 x64 | Windows/MSVC x64 | 68 verified |
These are maintainer-reported external results, not a local rerun. Each verifier pass includes manifest/digest validation, foreign decoding and byte-identical local re-encoding. The evidence covers the new position-distance archive and the unchanged 67-archive prefix. It does not cover other architectures or replace the public completion review. Local generation, 68-archive decode/re-encode equality, reordered-tail rejection and conversion checks for schemas 1 through 58 have passed.
The SHA-256 values detect accidental artifact changes but are not signatures and do not authenticate the producer. Use bundles downloaded from a trusted workflow run. GitHub may expire workflow artifacts according to repository retention settings; regenerate them by running CI for the required commit.
Schema 57 has local generation, exact-order verification, byte-identical
re-encoding, reordered-manifest rejection, and schemas 1 through 56
compatibility evidence. Its Windows/MSVC, Ubuntu 24.04/Ninja, and Ubuntu
26.04/Clang four-direction evidence is complete at revision
cdf90a4f93d3ef5c01db2c60a96bf6a439e02cd9. Schema 56's Windows/MSVC,
Ubuntu 24.04/Ninja, and Ubuntu
26.04/Clang four-direction evidence is complete at revision
c6bb7a62c7bfdcf6eef157ae1642f5de9576c182. Schema 55 did not receive a
standalone external bundle exchange; its frozen 65-archive prefix was verified
in every schema-56 bundle without assigning it a synthetic evidence record.
Schema 54's Windows/MSVC, Ubuntu 24.04/Ninja, and Ubuntu 26.04/Clang
four-direction evidence remains complete at revision
8ecc7a104c7b5def57737d9c8f9c40a63e6a8c30.
Schema 53's Windows/MSVC, Ubuntu 24.04/Ninja, and Ubuntu 26.04/Clang
four-direction evidence remains complete at revision
1de3df622106db7674bcf691201a601dae680294.
Interoperability work products are kept outside the source repository; only the resulting environment and verifier evidence are recorded here. These checks remain x86-64 evidence and do not cover a non-WSL Linux kernel.
Revision c4f831917a43f75ca5c698d19d3674f12803f40b received its first external
schema-7 cross-check on 2026-07-18. The external environment was Ubuntu 26.04
LTS under WSL2 on x86-64, using Ubuntu Clang 21.1.8, CMake 4.2.3, and PowerShell
7.6.3.
The Ubuntu 26.04 executable verified all eighteen archives from both the
Windows/MSVC and Ubuntu 24.04/Ninja CI artifacts, including byte-identical local
re-encoding. It then generated an ubuntu-26.04-ninja-x64 bundle. The local
Windows/MSVC executable independently verified all eighteen archives in that
bundle. Direct SHA-256 comparison across the three bundles found identical
input.bin bytes and identical bytes for every one of the eighteen archives.
This establishes deterministic x86-64 stream generation across MSVC and Clang and bidirectional decoding between Windows and the stated WSL2 Linux userland. It is historical schema-7 evidence.
Revision a4e3d1a5acb7bfc393aca4f2195188cfe0421817 received the corresponding
schema-8 cross-check on 2026-07-19. The external environment remained Ubuntu
26.04 under WSL2 on x86-64 with Linux kernel
6.18.33.2-microsoft-standard-WSL2, Ubuntu Clang 21.1.8, and CMake 4.2.3.
That executable verified all nineteen archives from the pushed Windows/MSVC and
Ubuntu 24.04/Ninja CI artifacts, generated an ubuntu-26.04-ninja-x64 schema-8
bundle, and verified all nineteen of its archives locally. The Windows/MSVC
executable then verified that Ubuntu 26.04 bundle in the reverse direction.
Every verification included exact local re-encoding, so the three producers
generated the same canonical archive bytes for every schema-8 profile.
Revision 8a854eaf9c7c6c36cc2d444cc8e1a135935887b2 received the schema-9
cross-check after its pushed CI completed successfully. The same Ubuntu 26.04
WSL2 x86-64 environment, using Ubuntu Clang 21.1.8, verified all twenty archives
from both the Windows/MSVC and Ubuntu 24.04/Ninja CI artifacts. It then
generated and verified an ubuntu-26.04-ninja-x64 twenty-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes required complete decode equality and byte-identical local re-encoding for every archive. This establishes canonical schema-9 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision bc8faba3043db78a953f18876f153abc847f814d received the schema-10
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8, verified all twenty-one archives
from both the Windows/MSVC and Ubuntu 24.04/Ninja artifacts. It then generated
and verified an ubuntu-26.04-ninja-x64 twenty-one-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes required complete decode equality and byte-identical local re-encoding for every archive. This establishes canonical schema-10 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 163948c61dd8b90359882bee122f16ab3794787c received the schema-11
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8, verified all twenty-two archives
from both the Windows/MSVC and Ubuntu 24.04/Ninja artifacts. It then generated
and verified an ubuntu-26.04-ninja-x64 twenty-two-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes required complete decode equality and byte-identical local re-encoding for every archive. This establishes canonical schema-11 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 7078d0ab20f6e0a1aeaa3c43e480ca866bf8a2fa received the schema-12
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8, verified all twenty-three
archives from both the Windows/MSVC and Ubuntu 24.04/Ninja artifacts. It then
generated and verified an ubuntu-26.04-ninja-x64 twenty-three-archive bundle.
The Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes required complete decode equality and byte-identical local re-encoding for every archive. This establishes canonical schema-12 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 77f16eaecfae20897f5d5f3e700584eb453fa3f1 received the schema-13
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8, verified all twenty-four
archives from both the Windows/MSVC and Ubuntu 24.04/Ninja artifacts. It then
generated and verified an ubuntu-26.04-ninja-x64 twenty-four-archive bundle.
The Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes required complete decode equality and byte-identical local re-encoding for every archive. This establishes canonical schema-13 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 802c7a1ab913b07ee79a04fa5b3390c061c88966 received the schema-14
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
twenty-five archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 twenty-five-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-14 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 504af4f6942aee7662bcb51abf9b55289c957d6c received the schema-15
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
twenty-six archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 twenty-six-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-15 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 01f746a5bef2225a0b8fa34f3ff9d52b42f13f40 received the schema-16
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
twenty-seven archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 twenty-seven-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-16 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision b4c700aca87fc925aab642cfb6a6b72f3a29c86b received the schema-17
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
twenty-eight archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 twenty-eight-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-17 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision fd11d1c7ef833873a02694da91f9f6d8d378948b received the schema-18
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
twenty-nine archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 twenty-nine-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-18 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision f8d51680a0ef827fa09f5782ad4ced4c335d346e received the schema-19
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and verified an
ubuntu-26.04-ninja-x64 thirty-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-19 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 01e87fe19f5c9c90edd87c9caeb8acf36b413aad received the schema-20
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
thirty-one archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and verified
an ubuntu-26.04-ninja-x64 thirty-one-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-20 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 110bf3c9f80f5bc3723232c6f027867e4c2e7a2f received the schema-21
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
thirty-two archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and
self-verified an ubuntu-26.04-ninja-x64 thirty-two-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-21 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 2aa51ded63bdeacb0e5b2ec28a21075a867bb353 received the schema-22
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
three archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-three-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-22 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 5397f261fa04ee49832d9f72b09960a156232aad received the schema-23
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
four archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-four-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-23 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision dad3638da2acb449afca969176194bf8323309f5 received the schema-24
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
five archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-five-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-24 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision bc4cfa45fc8787d5ec9277894bda0b10df0ef638 received the schema-25
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
six archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-six-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-25 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 5b2aa31ba3333c311ad4086b3438915a6c3ce36d received the schema-26
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
seven archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-seven-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-26 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision da376a7223f8a8072531271472f40d58b69e3b7a received the schema-27
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
eight archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-eight-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-27 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 3d5001ce7536c425328a597240244551605e8935 received the schema-28
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all thirty-
nine archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu
24.04 default-compiler/Ninja artifacts. It then generated and self-verified
an ubuntu-26.04-ninja-x64 thirty-nine-archive bundle. The Windows/MSVC
executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-28 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 2dcc17c09477958c1f8777a266ecfefbb75217d2 received the schema-29
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all forty
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 forty-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-29 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 827ddf085efb40c7d8f9bc27628977053179d84c received the schema-30
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all forty-one
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 forty-one-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-30 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 903181080556c3bb511ad4a2e5275837ebda48e7 received the schema-31
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all forty-two
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 forty-two-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-31 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision e9cf0c7d649cf32c9bc3a49bf3db9150370db381 received the schema-32
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
forty-three archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and
self-verified an ubuntu-26.04-ninja-x64 forty-three-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-32 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 2c30be4da1a80d01103dac0ee82fb0c4889f3af4 received the schema-33
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
forty-four archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and
self-verified an ubuntu-26.04-ninja-x64 forty-four-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-33 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 4929252144e4bfe44fb3ec076f548aa47e4ff111 received the schema-34
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all
forty-five archives from both the Windows/MSVC via Visual Studio 2026 and
Ubuntu 24.04 default-compiler/Ninja artifacts. It then generated and
self-verified an ubuntu-26.04-ninja-x64 forty-five-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-34 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 7c276151ab428aa9ba0376f8d9ba9a85a9fbd347 received the schema-35
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 46
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 46-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-35 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision bdcabd439d9cedb9e58f3dd2a3ac4dcb3526e1a2 received the schema-36
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 47
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 47-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-36 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 58b829dafa078e7dadd46e5de9ed7b1af45b5cc2 received the schema-37
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 47
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 47-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-37 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision b13cb7a51c782a66e63c493a7e5d1a5721edd86c received the project-version
0.3.0 release-candidate cross-check after its pushed CI completed
successfully. The Ubuntu 26.04 WSL2 x86-64 environment, using Ubuntu Clang
21.1.8 via Ninja, verified all 47 archives from both the Windows/MSVC via
Visual Studio 2026 and Ubuntu 24.04 default-compiler/Ninja artifacts. It then
generated and self-verified an ubuntu-26.04-ninja-x64 47-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest-order, size, SHA-256, fixture-decode, and byte-identical local re-encoding checks for every archive. This reconfirms canonical schema-37 bytes after the format-neutral HashChain Exact encoder promotion and establishes bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments for the 0.3.0 release candidate.
Revision 363a385168fcfab27adfc8eea3e302129cf01b15 received the schema-38
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 48
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 48-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-38 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision be940789f90b084bdf87ddd315b50da3e32fda55 received the schema-39
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 49
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 49-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-39 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision e74473d1511990ed06ea43c739783d1c58daf065 received the schema-40
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 50
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 50-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-40 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision c3ea5f87784faaca8c93e98fe5e459df3290747c received the schema-41
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 51
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 51-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-41 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision f64259a88c94adfed8fe590f308307c2f1d029aa received the schema-42
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 52
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 52-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-42 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision a871a05dad68c99712ba210a0b8a9f5f6eb6b3b3 received the schema-43
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 53
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 53-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-43 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 7f2b893f6ebeb968cab287420caa97201ce5b3b8 received the schema-44
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 54
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 54-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-44 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 2d5d582e974442a33151d0593c532e426e536e46 received the schema-45
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 55
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 55-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-45 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision e0ec10aef942c3ef9646b2332cec504ac176d4bf received the schema-46
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 56
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 56-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-46 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision c107daf9be319523402b0a4b4d9089c46a702ced received the schema-47
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 57
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 57-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-47 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision b030c0a63f4d8195c9904546e95e2122e3cac487 received the project-version
0.4.0 release-candidate cross-check after its pushed CI completed
successfully. The Ubuntu 26.04 WSL2 x86-64 environment, using Ubuntu Clang
21.1.8 via Ninja, verified all 57 archives from both the Windows/MSVC via
Visual Studio 2026 and Ubuntu 24.04 default-compiler/Ninja artifacts. It then
generated and self-verified an ubuntu-26.04-ninja-x64 57-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This reconfirms canonical schema-47 bytes after the format-neutral profile-helper consolidation and establishes bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments for the 0.4.0 release candidate.
Revision 891cadf5d0618b69f7993e4db47fb19da7da5f3e received the schema-48
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 58
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 58-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-48 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 0013741df87db1456a0deca0d7fc3345b3a1036a received the schema-49
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 59
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 59-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-49 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 21fa51568b5ab08ca37af0aa264ebbf39e9d7021 received the schema-50
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 60
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 60-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-50 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision f1d7517afdbd4b9f9bd8d5858390c868d1c4e5a9 received the schema-51
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 61
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 61-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-51 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 9b4b7250518cf39f1c25fd6dd29b18768e3557e4 received the schema-52
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 62
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 62-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-52 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 42783b435976777c40259ab5e90b616146535854 received the project-version
0.5.0 release-candidate cross-check after its pushed CI completed
successfully. The Ubuntu 26.04 WSL2 x86-64 environment, using Ubuntu Clang
21.1.8 via Ninja, verified all 62 archives from both the Windows/MSVC via
Visual Studio 2026 and Ubuntu 24.04 default-compiler/Ninja artifacts. It then
generated and self-verified an ubuntu-26.04-ninja-x64 62-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This reconfirms canonical schema-52 bytes after public BinaryTree Exact adoption and establishes bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments for the 0.5.0 release candidate. BinaryTree remains encoder-local, so this repeat changes no archive inventory or serialized identity.
Revision 1de3df622106db7674bcf691201a601dae680294 received the schema-53
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 63
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 63-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-53 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision 8ecc7a104c7b5def57737d9c8f9c40a63e6a8c30 received the schema-54
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 64
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 64-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-54 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision c6bb7a62c7bfdcf6eef157ae1642f5de9576c182 received the schema-56
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 66
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 66-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-56 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments. Schema 55 has no separate external record; its unchanged 65-archive prefix participated in all four schema-56 passes.
Revision a93b93e180d34fc3eda1be53f4dd03ae97336e39 received the schema-57
cross-check after its pushed CI completed successfully. The Ubuntu 26.04 WSL2
x86-64 environment, using Ubuntu Clang 21.1.8 via Ninja, verified all 67
archives from both the Windows/MSVC via Visual Studio 2026 and Ubuntu 24.04
default-compiler/Ninja artifacts. It then generated and self-verified an
ubuntu-26.04-ninja-x64 67-archive bundle. The Windows/MSVC executable
verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This establishes canonical schema-57 bytes across the three producers and bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments.
Revision cdf90a4f93d3ef5c01db2c60a96bf6a439e02cd9 received the project-version
0.6.0 release-candidate cross-check after its pushed CI completed
successfully. The Ubuntu 26.04 WSL2 x86-64 environment, using Ubuntu Clang
21.1.8 via Ninja, verified all 67 archives from both the Windows/MSVC via
Visual Studio 2026 and Ubuntu 24.04 default-compiler/Ninja artifacts. It then
generated and self-verified an ubuntu-26.04-ninja-x64 67-archive bundle. The
Windows/MSVC executable verified that bundle in the reverse direction.
Each of the four verifier passes reported the exact full revision and required manifest order, size, SHA-256, fixture decode, and byte-identical local re-encoding checks for every archive. This reconfirms canonical schema-57 bytes for the complete 64-MiB Contextual family and establishes bidirectional decoding between the recorded Windows and WSL2 Linux x86-64 environments for the 0.6.0 release candidate. The repeat changes no archive inventory or serialized identity.
Revision a141c160074eb061145e1a8f3439fcc83e6d16e7 received the
post-switch cross-check after the maintainer pushed it and reported successful
CI completion. The maintainer reported four successful verifier passes, each
with 67 archives and that exact full revision:
- Windows/MSVC Visual Studio 2026 CI bundle verified on Ubuntu 26.04.
- Ubuntu 24.04/Ninja CI bundle verified on Ubuntu 26.04.
- Ubuntu 26.04/Clang 21.1.8 bundle generated and self-verified there.
- The Ubuntu 26.04 bundle verified with Windows/MSVC in the reverse direction.
The verifier checks manifest order, size and SHA-256, decodes every archive, and compares local re-encoding byte for byte. The four reported results confirm the existing 67-archive schema-57 inventory across the three producer builds and both consumer environments for this revision. This repeat completes the external interoperability gate for the production HashChain best-length probe; it introduces no new codec ID, archive variant or schema revision.
The maintainer reported successful pushed CI for revision
be2febf16eff3e71cac00368ed975bf904ee4aa9 and four successful
schema-57 verifier passes, each reporting 67 archives and that same full
revision. On Ubuntu 26.04, the Windows/MSVC Visual Studio 2026 and Ubuntu
24.04/Ninja CI bundles both verified. A bundle generated with Ubuntu
26.04/Clang 21.1.8 self-verified there and then verified in the reverse
direction with Windows/MSVC.
The verifier checks the exact manifest order, sizes, SHA-256 values, foreign decode output, and byte-identical local re-encoding. This repeats the three-producer, two-consumer x86-64 exchange for the 0.7.0 release candidate without changing the schema-57 archive inventory or representation. These are maintainer-reported external results, not independently rerun here.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at 9b6e5df4d97471bfe4e0a231c92c6af1aac0b67b.
Each final verifier line reports 68 archives and that exact revision:
- The Windows CI bundle verified on the external consumer.
- The Ubuntu CI bundle verified on the external consumer.
- The externally generated bundle self-verified there.
- That external bundle verified in the reverse direction with the Windows build.
The verifier checks manifest order, sizes and SHA-256 values, exact decoded fixture bytes and byte-identical local re-encoding. These maintainer-reported results complete the external exchange gate for the private token scratch and grouped-literal decoder changes (DD-1295/DD-1296). They are not independently rerun results and do not extend to untested architectures or profiles. The schema-58 inventory, encoder representation and public ABI are unchanged.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at 3fb7c03488597b09a956463e1e88cf8a4cf5f595.
All four final lines report 68 archives and that exact revision. The Windows
and Ubuntu CI bundles verified on the external consumer; the externally
generated bundle self-verified there and verified in the reverse direction
with the Windows build.
The verifier checks manifest order, sizes and SHA-256 values, exact decoded fixture bytes and byte-identical local re-encoding. These maintainer-reported results close the external exchange gate for DD-1297's single-pass encoder scratch path, separately from the earlier decoder optimization exchange. They were not independently rerun here and do not extend to untested profiles or architectures. No archive inventory, format, public ABI or release change is introduced by this evidence record.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at 9a96f00b5025f07e5977fece692001a437658476.
Every final line reports 68 archives and that exact revision. Under the
established exchange workflow, the Windows and Ubuntu CI bundles verified
externally, the external bundle self-verified, and that bundle verified in
the reverse direction with the Windows build.
These maintainer-reported results close the external exchange gate for DD-1298/DD-1299's private single-pass entropy payload scratch. The verifier checks ordered inventory, sizes, SHA-256, exact decoded fixtures and byte-identical local re-encoding. Results were not independently rerun here. This record changes no archive inventory, format, public ABI or release state and makes no claim for untested architectures or profiles.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at 37b17a3e11ebac67fcfedb0b872c1675136f36e9.
All four final lines report 68 archives and that exact revision. Under the
established exchange workflow, the Windows and Ubuntu CI bundles verified
externally, the external bundle self-verified, and that bundle verified in
the reverse direction with the Windows build.
These maintainer-reported results close the external exchange gate for DD-1300/DD-1301's short-prefix comparison-order optimization. The verifier checks ordered inventory, sizes, SHA-256, exact decoded fixtures and byte-identical local re-encoding. Results were not independently rerun here. No archive inventory, format, public ABI or release state changes; these reports do not establish results for untested architectures or profiles.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at 0bbe882e2697aff8806c64b2ff4b5787ea6694ec.
Every final line reports 68 archives and that exact revision. Under the
established exchange workflow, the Windows and Ubuntu CI bundles verified
externally, the external bundle self-verified, and that bundle verified in
the reverse direction with the Windows build.
These maintainer-reported results close DD-1305's external exchange gate for compact dual-prefix production search. Earlier local verification passed 3,953 tests in each compiler build and both 68-archive exchange directions. The verifier checks ordered inventory, sizes, SHA-256, exact decoded fixtures and byte-identical local re-encoding. External results were not independently rerun here. No archive inventory, format, public ABI or release state changes; the reports do not establish results for untested architectures or profiles.
DD-1310 integrates an encoder-only removal of storage checks already proven by grammar acceptance. The checked reference remains available, the archive inventory remains schema 58 with 68 archives, and no stream format or public ABI changes. Both compiler builds pass all 3,954 tests; corpus payloads and fourteen CLI round trips remain byte-identical to retained archives.
Run the two local compiler exchange directions at the clean implementation revision, then obtain hosted CI and external exchange results naming that revision before closing its external gate. The previously reported results at 0bbe882e validate the earlier compact-prefix integration only. At this recording point, the new hosted/external gate is pending; no prior report is relabelled as verification of this change.
The maintainer reported successful pushed CI and four successful schema-58
verifier passes at f65cb552db9c0d8a04ddcc32991727af278b3272.
All four final lines identify 68 archives and that exact implementation
revision: one Windows CI bundle, one Ubuntu CI bundle and two passes naming
the external Ubuntu bundle. The repeated external bundle label is not a
fourth platform. These are maintainer-reported results, not independently
rerun hosted or external checks.
These results close DD-1310's CI/external exchange gate. Local validation at the same implementation revision also passed both 68-archive compiler exchange directions, following the two 3,954-test suites and bounded fuzz campaign. The verifier checks inventory, sizes, SHA-256, decoded fixtures and byte-identical local re-encoding. No format, public ABI, archive inventory or release state changes; no claim extends to untested architectures or profiles.
This entry supersedes the earlier current-inventory descriptions. New bundles
use schema 59 and marc-cli-v59, containing 69 archives. The first 68 entries
retain schema 58's order and representations; the last entry is
lzss-position-distance-dynamic-range-1m with identity 2/9 + 1/10 + 3/2.
The verifier continues to accept schemas 1 through 58 using their frozen lists.
Schema 59 requires the exact new codec set, archive count and order, hashes,
decoded fixture and byte-identical local re-encoding.
The common 8,193-byte fixture remains unchanged. It verifies exchange of this profile but does not exercise its full window: permanent tests separately cover the full frame and wide distances. Do not infer full-window coverage from the bundle fixture alone.
Run both local compiler exchange directions at the clean implementation revision, then push and obtain hosted CI and the four external verification results naming that revision. Successful new verifier lines must name 69 archives. Earlier 68-archive reports do not close this gate; repeated bundle labels do not establish additional platforms. Hosted/external qualification for schema 59 remains pending until the maintainer supplies those results.
The maintainer reported successful pushed CI and four successful verifier
passes at 1d351cc47895bc9bcc5dd856804b27068d675121. Each final line
identifies 69 archives and that exact revision: one Windows CI bundle, one
Ubuntu CI bundle and two passes naming the external Ubuntu bundle. The
repeated bundle label does not establish a fourth platform. These results
are maintainer-reported and were not independently rerun here.
These reports close the schema-59 CI/external exchange gate. Local validation also passed both 69-archive compiler exchange directions and all 4,019 tests in each compiler configuration; the frozen schema-58 prefix retained identical archive bytes. The 8,193-byte exchange fixture limitation above still applies. This record changes no format, public ABI, default, archive inventory or release state, and makes no claim for untested architectures or inputs.
The maintainer reported successful pushed CI and four successful verifier passes at revision 268398a263766bfa2ef40ad833a495b7fe63fb71. Every final line identifies 69 archives and that exact revision: one Windows bundle, one Ubuntu bundle and two passes naming the Ubuntu Clang bundle. The repeated bundle label does not establish a fourth platform. These are maintainer-reported results, not independently rerun hosted or external checks.
These reports close DD-1326's CI/external exchange gate for the public 1 MiB five-prefix encoder. At the same implementation revision, local validation passed both 69-archive compiler exchange directions with all archive hashes unchanged from the earlier schema-59 baseline. Both compiler regression suites passed 4,035 tests; the public sanitizer suite passed 20 tests, and all twelve verified corpus members retained their previous archive hashes and reconstructed exactly. Those local results remain distinct from the reported external passes.
Schema 59, its 69-archive inventory and the 8,193-byte exchange fixture remain unchanged. Full-window coverage comes from separate permanent tests, not from the bundle fixture. This qualification record changes no format, public ABI, defaults or release state and makes no claim for untested architectures or inputs.
The maintainer clarified the four IX-0052 results in their reported order:
| Pass | Archive producer | Verification consumer |
|---|---|---|
| 1 | Windows CI | Ubuntu |
| 2 | Ubuntu CI | Ubuntu |
| 3 | External Ubuntu Clang | Same Ubuntu environment (self-verification) |
| 4 | External Ubuntu Clang | Windows after transferring the bundle |
Verifier labels identify bundle producers, not the systems executing verification. The two equal Ubuntu Clang labels therefore represent self-verification and cross-system verification, respectively. This refines the evidence provenance; the reported revision, archive count and successful gate closure are unchanged.
The maintainer reported successful pushed CI and four successful verifier passes
at revision 2511312625e2c324b75ecda04eac4b1c3fe9263c. Each final line identifies
69 archives and that exact revision. The producer and consumer roles are:
| Pass | Archive producer | Verification consumer |
|---|---|---|
| 1 | Windows CI | Ubuntu |
| 2 | Ubuntu CI | Ubuntu |
| 3 | External Ubuntu Clang | Same Ubuntu environment (self-verification) |
| 4 | External Ubuntu Clang | Windows after transferring the bundle |
The two equal Ubuntu Clang labels identify the same bundle producer and distinct verification routes, not additional platforms. These are maintainer-reported results; hosted CI and external runs were not independently rerun here.
These reports close DD-1354's revision-specific CI/external exchange gate for the public prepared-mapping encoder. TVG-1220's local compiler, sanitizer, failed- frame and twelve-member CLI comparisons remain separate supporting evidence. Schema 59, its 69-archive inventory and the 8,193-byte exchange fixture remain unchanged; separate permanent tests cover full-window behavior. This record changes no implementation, format, ABI, defaults or release state and makes no claim about untested inputs or architectures.
The maintainer reported successful pushed CI and four successful verifier passes
at revision 1099a84be8dae4b1b821e7ccef2cd3ad58272546. Each final line identifies
69 archives and that exact revision. The producer and consumer roles are:
| Pass | Archive producer | Verification consumer |
|---|---|---|
| 1 | Windows CI | Ubuntu |
| 2 | Ubuntu CI | Ubuntu |
| 3 | External Ubuntu Clang | Same Ubuntu environment (self-verification) |
| 4 | External Ubuntu Clang | Windows after transferring the bundle |
The equal Ubuntu Clang labels identify the same producer and different consumer routes. These are maintainer-reported results; hosted CI and external runs were not independently rerun here.
These reports close DD-1359's revision-specific CI/external exchange gate for the public finder-scratch encoder. TVG-1226's local full suites, sanitizers, failed-frame tests and twelve-member CLI comparisons remain separate evidence. Schema 59, its 69-archive inventory and the 8,193-byte exchange fixture remain unchanged; full-window coverage comes from separate permanent tests. This record changes no implementation, format, ABI, defaults or release state and makes no claim about untested inputs or architectures.
Current bundles use schema 60 and marc-cli-v60, containing 70 archives.
Entries 1 through 69 retain the complete schema-59 order and byte streams;
entry 70 is lzss-position-distance-dynamic-range-4m with exact Format 2.0
identity 2/10 + 1/11 + 3/2. The generator validates this identity before recording
the archive. The verifier independently checks its bounded header even when
the manifest hash matches, then restores and canonically re-encodes the bytes.
The shared fixture remains 8193 bytes and does not exercise its full window.
TVG-1254 qualifies both local compiler routes, schemas 1 through 60, reordered
inventory and rehashed crossed-identity negatives, unchanged old 69 archive
bytes/hashes, and opposite-compiler consumption of all 70 entries. The new
four-MiB archive is 1247 bytes, SHA-256
7b2e83edb38ff40b6402cb5bb90c577bb3bd81b4869337324b040e51d765365e.
DD-1386 supplies separate complete-corpus/full-window and native-suite evidence.
Existing hosted workflows call the updated generator without changing their producer roles. Hosted CI and these four maintainer external routes remain pending for the actual schema-60 revision:
| Route | Producer | Consumer |
|---|---|---|
| 1 | Windows CI | Ubuntu |
| 2 | Ubuntu CI | Ubuntu |
| 3 | External Ubuntu Clang | Same Ubuntu environment |
| 4 | Same external Ubuntu Clang bundle | Windows after transfer |
Earlier 69-archive reports do not qualify this new archive or revision. Local cross-compiler checks do not substitute for those external results.
On 2026-10-02 the maintainer reported four successful verification results for
revision 02916f6fbf5ab6544d42fe432d019f91d088c2f6, each identifying 70 archives.
Following the previously clarified producer/consumer roles, the results are:
| Route | Producer | Consumer | Verified archives |
|---|---|---|---|
| 1 | Windows CI | Ubuntu | 70 |
| 2 | Ubuntu CI | Ubuntu | 70 |
| 3 | External Ubuntu Clang | Same Ubuntu environment | 70 |
| 4 | Same external Ubuntu Clang bundle | Windows after transfer | 70 |
The last two equal producer labels represent one bundle and two consumers, rather than two independently produced bundles. These are maintainer-reported results; the external executions were not independently rerun here.
The reports close DD-1387's four-direction external exchange gate for schema 60 at that exact implementation revision. They do not assert full-window coverage from the 8193-byte fixture; DD-1386's corpus/window tests and DD-1387's local inventory/negative tests remain separate evidence. The maintainer subsequently confirmed successful GitHub CI completion for the same revision. Together these reports close DD-1387's revision-specific CI and external exchange gate. This evidence update changes no codec, format, ABI, limits, defaults, archive inventory or release state.
Current producers emit schema 61 and marc-cli-v61 with exactly 71 archives. Entries 1 through 70 preserve the frozen schema-60 order and byte streams; entry 71 is lzss-position-distance-dynamic-range-8m with exact Format 2.0 identity 2/11 + 1/12 + 3/2. The 8193-byte fixture and first 42 stable profiles are unchanged. Consumers retain schemas 1 through 60.
The verifier admits all entries before starting codec work. Directed malformed manifest tests reject downgrade/missing/duplicate/order/codec-set/hash/size and rehashed identity/header tampering with no decoded or re-encoded output. Local source-bound tools generate and verify identical complete bundles; historical schema compatibility and a frozen-prefix producer comparison pass. These local results do not establish maximum-window coverage from a small fixture.
Hosted CI and four external consumer roles remain pending for the actual schema-61 revision: consume each of the two hosted producer bundles, then consume one independently produced external bundle on its own platform and on the other platform. Repeated producer labels describe those two consumer roles. Earlier schema-60 receipts do not qualify this inventory or revision. No new external execution or release state is inferred here.
On 2026-10-05 the maintainer confirmed push and successful GitHub CI completion, and supplied four successful 71-archive verifier reports at revision e3a4ee916fd90687fa73614210a23c46b58e3ebe, including the DD-1474 watchdog correction.
The reported roles are consumption of each of the two CI-produced bundles, then consumption of one independently produced external bundle on its own platform and on the other consumer platform. The last two reports name the same producer and represent two consumer roles, not an additional producer. All reports refer to the same revision and complete schema-61 inventory.
These maintainer-reported results close DD-1473's revision-specific CI and four-direction external exchange gate at that revision. The executions were not independently rerun by the agent. The small 8193-byte fixture establishes exchange and deterministic re-encoding; maximum-window and failed-frame publication evidence remains in the separate prior qualifications. The earlier timeout report is superseded by the successful CI report for the corrected revision. No codec, format, ABI, limits, defaults, inventory or release state is changed by this evidence update.
Schema 62 and marc-cli-v62 append lzss-position-distance-dynamic-range-16m as archive 72, with exact identity 2/12 + 1/13 + 3/2. The unchanged prior producer and three freshly linked local production routes agree on every byte of the independent 8193-byte input and frozen first 71 archives. All 72 archives agree across the new routes and pass decode plus deterministic re-encode checks. Comparison fixtures use explicit synthetic metadata; deliverable manifests identify the actual integration revision and producer binary.
Genuine manifests for schemas 1..62 pass. Twenty-seven new negative cases include false downgrade, codec set, count/order/duplicate, hash/size, truncation and rehashed dictionary/context/version identity bytes. A guarded entry point proves rejection before any codec launch; no decoded/re-encoded file appears. Maximum-window and failed-frame publication remain separately qualified by actual CLI and public-core tests, rather than inferred from small bundles.
IX-0058's maintainer-reported CI and four external 71-archive results apply only to e3a4ee916fd90687fa73614210a23c46b58e3ebe. Hosted CI and four-direction 72-archive external exchange for the new integration revision have not yet been reported. The maintainer performs push and supplies those results; local qualification is not described as external verification.
On 2026-10-05 the maintainer confirmed push and successful GitHub CI completion, and supplied four successful 72-archive verifier reports at revision 8e8a77c55957ea334fbf3632f6e9c1d7500f32b4.
The reported roles are consumption of each of the two CI-produced bundles, then consumption of one independently produced external bundle on its own platform and on the other consumer platform. The last two reports identify the same producer and represent two consumer roles, not an additional producer. All four reports identify the same revision and complete schema-62 inventory.
These maintainer-reported results close DD-1487's revision-specific hosted CI and four-direction external exchange gate. They supersede IX-0059's pending status for that revision. The agent did not independently rerun those external executions or inspect the hosted CI run. The small 8193-byte exchange fixture does not replace the separately qualified maximum-window, memory-limit and failed-frame nonpublication tests. This evidence update changes no codec, format, ABI, limits, defaults, inventory or release state.
Schema 63 and marc-cli-v63 append lzss-position-distance-dynamic-range-32m
as archive 73 with exact identity 2/13 + 1/14 + 3/2. Preserve the schema-62
prefix of seventy-two archives and the independent 8193-byte fixture.
Qualify generation, decoding, deterministic re-encoding, all genuine schemas
1..63 and guarded whole-manifest rejection before any codec launch.
Maximum-window, resource and failed-frame behavior have separate evidence.
IX-0060's maintainer-reported CI and four external 72-archive results apply to 8e8a77c55957ea334fbf3632f6e9c1d7500f32b4. They are not evidence for the new 73-archive integration revision. Local qualification, final producer metadata, hosted CI and external reports are attributed separately; the maintainer performs push and provides revision-specific external results.
Local qualification for IX-0061 passes all 4,041 tests in the complete configured suite and three independently rebuilt production CLI routes. All seventy-three archives decode and deterministically re-encode; the entire old seventy-two-archive prefix is byte-identical to a separately frozen bundle from the prior qualified runtime. The three producer bundles have identical input and archive bytes. Preserved genuine schema inventories 1..63 all pass, including the historical name conversion. Each route rejects twenty-seven new schema-63 negatives before any codec launch or output publication; previous schema-specific guards remain in the complete suite. Instrumentation covers the whole production library with address/undefined-behavior checking and leak detection disabled. Synthetic fixture revision metadata is kept separate from final clean-source revision-bound producer bundles. Hosted CI and the four external consumer roles for the new revision remain maintainer-owned gates.
On 2026-10-06 the maintainer confirmed push and successful GitHub CI completion and supplied four successful 73-archive verifier reports, all at revision 8c5b4d00c9dcb487d4e1215754d4cd8456e6fc51.
The reported consumer roles are the CI-produced Windows bundle consumed on an external Ubuntu consumer, the CI-produced Ubuntu bundle consumed on that consumer, an independently produced external bundle self-consumed, and that same external bundle consumed on Windows. The last two identical producer labels identify two consumer roles for one producer; they are not counted as two independent producers. All four reports cover the schema-63 inventory.
These maintainer-reported results close DD-1499/IX-0061's revision-specific hosted CI and four-direction external exchange gate after the DD-1500 test- fixture correction. They supersede the pending external status for this revision. The agent did not inspect the hosted CI run independently or rerun these external executions. The 8193-byte exchange fixture complements the separate maximum-window, resource, allocation-failure and failed-frame nonpublication evidence; it does not establish a new performance result. This evidence update changes no codec, format, ABI, limits, defaults, inventory or release state.
Schema 64 / marc-cli-v64 appends only lzss-position-distance-dynamic-range-64m as the seventy-fourth archive after the unchanged seventy-three-profile prefix. Exact identity is Format 2.0 dictionary 2/14 + context 1/15 + entropy 3/2, with fifty contexts. DD-1510 requires a frozen prior byte comparison, genuine schema-1..64 inventories and whole-manifest prelaunch rejection, including rehashed low/high identity bytes and context count. Input remains the independent 8,193-byte recipe. Local source-bound producer/consumer and historical qualification is pending; future hosted CI and four maintainer external routes must be reported separately.
Local qualification for IX-0063 passes all 4,062 tests in the complete configured suite. Three independently rebuilt production CLI routes generate identical input and all seventy-four archive bytes. Each self-verifies and all six distinct producer/consumer pairs restore and deterministically re-encode. Every old seventy-three archive byte agrees with a separately frozen bundle from the prior qualified runtime. Genuine manifests for schemas 1..64 retain their actual inventories, including the historical name conversion. Each new producer route passes thirty schema-64 rejection cases with zero codec launches and zero output files; schema-63's unchanged twenty-seven cases also pass. These are local fixture observations with explicitly synthetic revisions; final clean-source production bindings and runtime publication are separate steps. Hosted CI and four external consumer reports remain maintainer-owned gates for the new revision. No external implementation was consulted.
On 2026-10-06 the maintainer confirmed push and successful GitHub CI completion and supplied four successful 74-archive verifier reports, all at revision 55e24b9b4be9642e507846a161e8638aeb8e3ce9.
The reported consumer roles are the CI-produced Windows bundle consumed on an external Ubuntu consumer, the CI-produced Ubuntu bundle consumed on that consumer, an independently produced external bundle self-consumed, and that same external bundle consumed on Windows. The last two identical producer labels identify two consumer roles for one producer, not two independent producers. All four reports cover schema 64 / marc-cli-v64, including lzss-position-distance-dynamic-range-64m as the seventy-fourth archive.
These maintainer-reported results close DD-1510/IX-0063's revision-specific hosted CI and four-route external exchange gate. They supersede the pending external status for this revision. The agent did not independently inspect the hosted CI run or rerun these external executions. The exchange fixture complements the separate full-window, resource, allocation-failure and failed-frame nonpublication evidence; it does not establish a new performance result. This evidence update changes no codec, format, ABI, limits, defaults, inventory or release state. Reported CI success applies to the revision above, not to a subsequent documentation-only commit.
Schema 65 / marc-cli-v65 contains 75 archives. Entry 75 is
lzss-position-rans-1m, with exact 2/9 + 1/10 + 4/4 identity, log 12,
scalar-state identifier 1, 44 contexts and 2566 frequency entries. The input
recipe and first 74 entries remain unchanged. The new archive is 3014 bytes,
SHA-256 d17bf0e8491c62e3003489a97e089d702ea3e69f64a81a47b204275d3077dca8.
This small fixture does not exercise the complete one-MiB window; separate
full-corpus, frame-boundary and malformed-lifecycle evidence remains required.
Three independently built local CLI producers generate identical input and all 75 complete archive bytes. Each self-verifies and all six distinct cross pairs decode and deterministically re-encode every archive. All prior 74 archives equal a separately frozen schema-64 production bundle. Each producer route rejects all 36 new negatives before any codec launch or output creation. These local fixture manifests use explicitly synthetic revisions. Final commit bindings and fixed-runtime publication are separate steps. Hosted CI and four maintainer external routes for the new revision remain pending; old schema-64 reports do not cover this new archive.
Schema 65 (marc-cli-v65) retains its 75-archive inventory and its first
74 entries. Entry 75 and its filename are corrected to
lzss-position-distance-rans-1m and
lzss-position-distance-rans-1m.marc before publication. Archive bytes are
unchanged. The earlier unpublished spelling is rejected during manifest
admission, before launching a codec. Schemas 1 through 64 are unchanged.
Date: 2026-10-07. Verified revision:
c17ed4351160eff3f9046d43d0fa9b60bac32f34.
The maintainer reports hosted CI success and four successful external routes,
each verifying all 75 archives: CI-produced Windows archives consumed by
Ubuntu, CI-produced Ubuntu archives consumed by Ubuntu, an independently
produced Ubuntu bundle consumed by its producer, and that Ubuntu bundle
consumed by Windows. The two final reports carry the same producer label
because they consume the same bundle through different execution routes.
These reports cover schema 65 / marc-cli-v65, with the canonical
lzss-position-distance-rans-1m archive as entry 75.
The agent inspected the maintainer-supplied CI log archive: both main jobs pass all 4211 tests; all four installed static/shared package jobs pass both consumer tests. Checkout records bind all six jobs to the revision above. The agent did not rerun or independently observe the four external routes. This closes the revision-specific hosted CI and external exchange gates previously pending in IX-0065/IX-0066. BM-0217 records observed CI durations. It changes no codec, format, ABI, limits, defaults, inventory or release state. This success does not apply automatically to a later documentation commit.
Schema66 (marc-cli-v66) appends only lzss-position-distance-rans as
archive76 after the frozen schema65 order. Dictionary/context/entropy identity
is 2/8 + 1/9 + 4/5, frame/window 65,536 bytes, forty contexts and 2522
frequencies. Generation checks exact identity and immediate roundtrip;
verification checks every stream-header byte, including original-size binding
and reserved zeros, before codec launch. Old schemas1..65 remain accepted.
Compatibility reconstructs schema65 by removing only archive76, then follows the unchanged historical chain. All first75 native archive bytes match the frozen bundle. The complete schema1..66 suite passes, including125 new manifest/inventory/hash/size/identity/header negatives that launch no codec and publish no files. New negative cases reuse one private fixture and retain each changed manifest/archive, preserving source and prior archives. Actual revision-bound bundles and maintainer-hosted CI/external reports must be recorded separately; older external successes do not qualify this schema.
Date: 2026-10-08. Verified revision:
15a8fae6872947191d19718ae056f7f07f5c9001.
The maintainer reports a successful push and hosted CI, followed by four
successful external routes, each verifying all 76 archives: CI-produced
Windows and default Ubuntu compiler bundles consumed by Ubuntu, and an
independently produced Ubuntu Clang bundle consumed by its producer and by
Windows. The final two reports share a producer label because their consumers
differ. They are four route results covering schema66 / marc-cli-v66,
including lzss-position-distance-rans as archive76.
These maintainer reports close the revision-specific hosted CI and external exchange gates left pending in IX-0068. The agent checked the reported revision against its existing qualification receipt and unchanged sources/runtime. No CI log archive was supplied for this report; the agent did not independently inspect CI test counts or durations, or rerun the four external routes. No codec, format, ABI, default, limit or inventory changes. This result does not automatically qualify a later revision.
Schema67 and marc-cli-v67 append archive77 for the case-sensitive selector lzss-position-distance-rans-4m. The first76 entries retain their order and representations. Require exact tuple2/10+1/16+4/7,54 contexts,4636 frequencies, 4194304-byte frame/window and known original size across all112 header bytes before launching a decoder. Unknown names, duplicate/order/hash/size errors, identity corruption and truncated headers must publish no output and launch no codec. Old schemas retain their respective inventories. Local producer/cross-path and historical qualification is IX-0070.
Date: 2026-10-08. Implementation revision
994587b609abc8074ace0d4a16907fa428cb4c62.
Two compiler implementations produce and self-verify77 archives; each
consumer also decodes the other producer's complete bundle. All77 archive
bytes match across compilers. Both producers preserve every byte of the
first76 entries in the frozen schema66 bundles.
Both producers'125 new negative cases reject before codec launch, with no output publication. The complete schema1..67 conversion/verification suite passes, including the existing historical negative checks. That synthetic history suite uses its explicit zero revision marker; it is distinct from the actual implementation-revision bundles above. All earlier artifacts are retained. Hosted CI and maintainer-run external routes for this new revision remain unreported; previous schema66 successes do not qualify it.
Date: 2026-10-08. Verified revision:
80fb9be34bf50c565b54185d5d3f7655e07d2b61.
The maintainer reports a successful push and hosted CI, followed by four successful external routes, each verifying all77 archives: CI-produced Windows and default Ubuntu compiler bundles consumed by Ubuntu, and an independently produced Ubuntu Clang bundle consumed by its producer and by Windows. The duplicate Clang producer label represents two consumers. These reports cover schema67/marc-cli-v67, including archive77 lzss-position-distance-rans-4m.
The reports close the revision-specific hosted CI and external gates after IX-0070. The agent matched the reported revision and unchanged runtime to the existing local completion receipt. No CI log archive was supplied for this report; CI counts/durations were not independently inspected, and the external routes were not rerun by the agent. No codec, representation, limit or inventory changes. This does not qualify a later revision.
Schema68 and marc-cli-v68 append archive78 for the case-sensitive selector lzss-position-distance-rans-8m. The first77 entries retain their order and representations. Require exact tuple2/11+1/17+4/8,55 contexts,4647 frequencies, 8388608-byte frame/window and known original size across all112 header bytes before launching a decoder. Unknown names, duplicate/order/hash/size errors, identity corruption and truncated headers must publish no output and launch no codec. Old schemas retain their respective inventories. Two local compiler producers pass two self/two cross routes, all78 archive bytes agree, first77 equal the frozen schema67 controls, and both125 preflight refusal campaigns and complete history1..68 pass. Final source-bound metadata and fixed-runtime qualification follow this milestone; hosted CI and external routes for the final81 inventory are not claimed.
Date: 2026-10-08. Two local compiler producers verify78 archives each on self and cross consumers; every emitted archive matches across compilers. All first77 archives equal frozen schema67 bytes. Both125 strict-header negative campaigns reject before codec launch and publish no output. Synthetic zero-revision history1..68 passes, including retained earlier refusal cases. Source-hashed exchange scripts are exercised with the f63cf327 public runtime snapshot; final committed-source metadata and fixed runtime replacement follow. No local third compiler family, new hosted CI or maintainer external validation is claimed. The remaining16/32/64MiB and final81-archive gates are pending.
Date: 2026-10-08. Fresh schema68 bundles bind implementation revision 6f7987f104695fafaeb14e4c19b9465686eae653 and the actual qualified executables. Two compiler producers pass both self and cross consumers; all78 archives match and the first77 remain identical to the frozen schema67 controls. The final fixed-path CLI passes the public lifecycle/static/shared ABI/CLI targets and all twelve complete corpus members, matching the independent candidate and the other compiler's public CLI byte for byte. Previous runtime files are retained and their copied hashes checked. IX-0072's125 refusal cases per producer and history1..68 remain qualified; no new hosted CI or maintainer external result is claimed. Eight-MiB local qualification is complete. The16/32/64MiB profiles and final81-archive audit remain pending, so the full four-window goal stays active. Local milestone commits do not require completion of the other windows; push does.
Schema69 and marc-cli-v69 append archive79 for the case-sensitive selector lzss-position-distance-rans-16m. The first78 entries retain their order and representations. Require exact tuple2/12+1/18+4/9,56 contexts,4658 frequencies, 16777216-byte frame/window and known original size across all112 header bytes before decoder launch. Reject unknown names, duplicates, order/hash/size errors, identity corruption and truncated headers without codec launch or output publication. Earlier schemas retain their respective inventories. Final committed-source exchange and fixed-runtime qualification are separate gates; no new hosted CI or maintainer external result is claimed here.
Date: 2026-10-08. Two local compiler producers pass two self and two cross consumers for79 archives. All archive bytes agree; the frozen first78 remain exact. Both125 preflight refusal campaigns reject before codec launch and publish no output. Synthetic zero-revision history1..69 and earlier refusal cases pass. Source-hashed exchange scripts are exercised with the022e294b public runtime snapshot. Final committed-source metadata, fixed-runtime replacement and isolated directional/peak/default admission remain pending. No new hosted CI or maintainer external result is claimed. The32/64MiB and final81-archive gates remain pending; commit this passing exchange milestone locally without waiting for the other profiles.
Date: 2026-10-09. The qualified current public runtimes reproduce all twelve corpus archives byte for byte and restore their exact raw inputs. Schema69 bundles at implementation revision3ed53384fff1baeda7ac95e746679f0a0fd6e300 pass two local self and two local cross routes for79 archives. All79 producer outputs agree and the first78 remain equal to the qualified schema68 inventory. Exchange script hashes and archive bytes remain equal to IX-0074's125-refusal and history1..69 evidence; those records are retained. Public API/static and shared ABI/CLI tests pass with the current runtimes. BM-0224's48 isolated directional rows, resource/default review and five fully instrumented10000-run fuzz campaigns complete the sixteen-MiB local gates. Earlier artifacts are preserved. Initial exchange script startup failures occurred before codec launch and were corrected; successful retries are recorded separately. No hosted CI or maintainer external result for this revision is claimed. The32/64MiB profiles and final81-archive audit remain pending. Passing milestones are committed locally; completing all four profiles remains a condition before push.
Date: 2026-10-09. Schema70/codec-set marc-cli-v70 appends one archive for lzss-position-distance-rans-32m, producing80 archives. Preserve all first79 entries in schema69 order and exact bytes, including the earlier frozen77. The appended112-byte stream header must carry exact2/13+1/19+4/10, frame/window33554432,57 contexts and4669 normalized frequency entries. Construct an independently specified complete header and compare all112 bytes, including reserved bytes and declared original size, before codec launch. The rest of the manifest admission rules remain unchanged.
Two local producer/compiler bundles agree in all80 archive bytes and retain the qualified schema69 prefix. Two self and two cross consumers pass. Each producer's125 negative cases reject before codec launch or output publication, including mutation of each header byte. The synthetic revision history test verifies schemas1 through70 and previous negative fixtures; synthetic history is not an external producer/revision claim. Full32MiB corpus/isolated measurements, public capacity admission and final committed-source/runtime qualification remain pending. This is a local milestone5 checkpoint.64MiB/final81 and maintainer push, hostedCI and external four-route checks remain separate unfinished gates.
The implementation at revision 7c66f0258f3980af43509f2c670065b30dc3708e passes the final full twelve-member corpus on both qualified runtimes, with archive bytes equal to the independent private results and every raw roundtrip verified. Public API, static C ABI, shared C ABI and CLI tests pass on both builds. Schema70 contains exactly80 archives; its first79 archives remain byte-for-byte unchanged. Both producer bundles pass two local self routes and two local cross routes with source/runtime bindings.
The earlier125 prelaunch negative cases per producer and synthetic history1..70 evidence are retained, not rerun: the tested schema scripts and every archive body are verified unchanged. BM-0227 supplies the full 48-row measurements and admits the unchanged1536MiB capacity default; FZ-0085 supplies five fully instrumented10000-run campaigns. Thirty-two-MiB is locally complete. Sixty-four-MiB, final81-archive qualification and maintainer push/hosted CI/external four-route validation remain pending.
Schema71/codec-set marc-cli-v71 appends lzss-position-distance-rans-64m
as archive81. Its first80 entries retain schema70 order, including the frozen
first77. The verifier retains schema1 through70 dispatch and checks the new
112-byte stream header before invoking any codec. The new identity is
dictionary2/14, entropy4/11, context1/20, frame/window67108864 bytes,
58 contexts and4680 frequencies; all other bytes retain the exact canonical
header rules, including the manifest's original length.
The historical harness derives schema70 from the schema71 first80 prefix before continuing its earlier history checks. The new125-case manifest negative suite must reject every changed manifest/header before codec launch and output publication. This extension is prepared, not exchange-qualified: script parsing and the two identity helpers pass the actual valid header and 114 rejection cases each (112 changed bytes, truncation and wrong original length). These helper checks do not prove manifest admission, archive-prefix preservation, bundle generation or self/cross decoding. Full schema71 negative, history1..71, both producer bundles, two self and two cross routes and final source/runtime bindings remain pending. Maintainer push, hosted CI and external four-route validation remain separate gates.
The source at5515df5c300cec10f59ed9b1597d10f92293470d produces exactly81 archives on both normal public builds. Archive names, order, sizes, SHA-256 and actual bytes agree. All first80 archives remain byte-for-byte unchanged from the qualified schema70 bundle, including the frozen first77.
Two local self routes and two local cross routes restore every original fixture. Each producer's125 negative cases reject before codec launch and output publication. The historical harness verifies schemas1 through71, retaining its evidence. Synthetic history uses a synthetic revision and does not claim historical producer or external revision verification (TVG-1419). The full private/public twelve-member corpus also passes BM-0230.
This completes the local milestone5 exchange checkpoint. Isolated64MiB encode/decode timings, normal peaks and capacity admission remain unfinished, as do the fixed runtime update and final committed-source81 qualification. Maintainer push, hosted CI and external four-route checks remain pending; the whole four-profile goal remains active.
At implementation revision615be3e941fa069caa5e937d6d6a9d440494e094, both final normal builds pass thirteen public API, static/shared C ABI and CLI regressions covering the four large-window position-distance rANS profiles. Both producers generate identical81-archive schema71 bundles. Two self and two cross consumers restore all originals; each producer's125 negative cases rejects before codec launch or output publication. Every preceding80 archive body remains unchanged, including the frozen77 prefix.
The final audit rechecks retained8/16/32MiB closure evidence and unchanged profile-specific production sources, then combines the current shared regressions with all five64MiB milestones. It retains the qualified schema1..71 historical evidence without claiming a new historical runtime run. BM-0231 and FZ-0086 complete the measurement, capacity and fuzz gates. All four profiles are locally complete. Maintainer push, hosted CI and external four-route validation of all81 archives remain required before the overall goal is complete. This closure changes documentation only; the implementation revision and bundle source bindings remain explicit.