Skip to content

Forward merge release/v0.4 (0.4.0, 93fbe7df9) into release/v0.5 - #322

Merged
afflom merged 325 commits into
release/v0.5from
claude/forward-merge-v0.4-v0.5
Oct 26, 2026
Merged

afflom merged 325 commits into
release/v0.5from
claude/forward-merge-v0.4-v0.5

Conversation

@afflom

@afflom afflom commented Oct 26, 2026

Copy link
Copy Markdown
Member

The forward merge of release/v0.4 at 93fbe7df9 (the 0.4.0 cut, through #318) into this line (§9.4). Every row of the 0.4.0 register joins this line's register: 472 rows (433 base + 16 v0.5 + 23 v0.4). The hand-written sources are the union of both sides; the generated files were taken from release/v0.4 and then regenerated on the merged tree.

Two identifier collisions, resolved by the rule that the release/v0.4 row keeps its number:

Other resolutions. The bootstrap validator RustValidate and the Rust oracle carry both #279's loop-end tail check and H4b's machine callee and overflow value; §17.13 narrows the loop-end tail to a call ("The operand of an Err that ends a loop … is in tail position when it is a call"). migrate.rs uses the compiler's own Sha256Digest since this line no longer ships sha2. Workspace version 0.4.0. SPEC.md was merged by hand and spec/src rebuilt to project it byte for byte.

Regeneration. The renderer record moved with the merged runtime and the language-1.2 compiler-semantics ID with it: fc363733… → bf827c3cd136d83b2bd6a7ddcfc44d53b6331ea86033ba188146b9d83d2e6df7 (1.1 373f6133… and 1.0 c7a28ab1… unchanged). Locks, goldens, calculus, production export, fixtures, provenance, generated documents and the verification records of every project but the Atlas are rewritten; the Atlas records follow from this PR's first gate run. Two cases needed the merged tree's rules: the spec sources are formatted (NS-01), and SL-07 copies an example that imports a source package beside the workspace file (LLC0108 reached this line with the merge).

Checks on the merged tree: validate-model (472 ids, audits clean), validate-spec-links (472 bijective), fmt, BS-05 with the three-way constants, and rb_07 rp_10 rp_01 rp_07 rp_11 tc_07 ns_01..05 sl_06..08 rb_22 pass.

After this lands: #299 retargets to release/v0.5; #303, #295, #249, #239 proceed; a second forward merge brings #298 (H5) and #320/#321.


Generated by Claude Code

Alexander Flom and others added 30 commits October 20, 2026 15:24
…w) as the Rust validator does (isOverflowErr, three crates, counts 442/195/305/81 = 747); core_from, so the in-case plants go through the call core() makes, with the bare-Ok plant; SPEC.md 17.16 fixed fuel, match on a block, both validators; records
…isOverflowTy, BS-05 counts with i128 (443/195/306/82 = 749), SP-22 in a bdd part of its own with the plant on one theorem, identity_bytes reads on a stack of its own
…readers; the measurement and the single identity regeneration recorded
…ed locally; the bare-Ok plant names the shared arm, not nat_add
…nd the meaning of equal and compare: register rows, SPEC.md 31 rows, scenarios, and the failing named tests
…y point, so a second data module reads through the same closed tables
…pport items of the core runtime as its data, shipped byte for byte
… validator with its named rules, and the core runtime's base replaced by the printed support crate; RB-17's case
…two sets, SPEC 17.16 and the register restated
…first scalar, so each elaborates within the host's memory
…ith the natural and the integer, a CI part for the support leg
… decodes once, the refines_<path> clause in SPEC, VERIFICATION's SP-22 command and the 142 s label
…er nested keys of depth two and three, measured
std-models, std-reasoning, and the helpdesk application (#32, L2.3)
…ne: BS-05's package counts follow the three new committed programs (226 crates; 432 validateCrate and 295 correspondCrate vectors, to be measured)
…: the register at 439, BS-05's counts follow the three new committed programs
claude and others added 23 commits October 24, 2026 22:49
… but the Atlas, on the moved identity

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…ppend, the one line append_list now calls

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…teen shard artifacts of the first gate run

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…ap-h4a

#91 stage 4, H4a: the first four std runtime functions as source — append_list, length_list, index_list, slice_list (SP-29, RB-20)
…emantics identity; locks, goldens, calculus, packages, fixtures and provenance regenerated

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
… but the Atlas, on the moved identity

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…teen shard artifacts of the first gate run

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…ap-h4b

#91 stage 4, H4b: the byte and text functions of the std runtime as source; landed with its identity regeneration

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…34-version-0.4.0

The line gained #293 and #297 and regenerated the 25 build manifests and the
two fixture records on the moved compiler-semantics identity; those 27 files
are taken from the line here and rewritten for 0.4.0 in the next commit. The
register is 456 at every site (the bump changes no count).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… moved identity

compiler.version in the 25 expected/build/manifest.json files and the two
fixture records that hash them (cl-05-build-artifacts, vr-11-exact-success),
written with cargo xtask check-golden --write and check-fixtures --write over
the regenerated records of the line; the manifests differ from the line's only
in the version string.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…he register, the three compiler-semantics IDs, and exactly which bootstrap and runtime stages this tag ships

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…nly at this tag, and the language-model rows it carries

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…reads prose too)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
….4.0

[0.4] #34: the workspace version is 0.4.0 and the 0.4.0 release notes; the v0.4.0 tag commit

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…ts, so the release run starts

GitHub validates a reusable workflow at startup against the permissions and
secrets its caller passes, before any job condition is read. crate.yml's
upload job declares id-token: write and reads the crates.io secret, and the
release workflow granted neither, so every release run, rehearsal or tag,
ended as a startup failure (run 40219629854 on release/v0.4 at 71c6bd8).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…e cargo version used

cargo 1.97.1's `publish --dry-run` leaves the archive under
`<target-dir>/package/tmp-crate/`, where the readiness step read
`<target-dir>/package/<crate>` and failed with "No such file" (rehearsal
40220072633, job crate readiness). Reproduced locally with the pinned
toolchain; the archive's bytes equal `cargo package`'s, as the step asserts.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
… name too, and the oracle reads the lines the workflows now carry

The publish job of release.yml read the dry-run archive at the path cargo
1.97.1 no longer uses, as the crate readiness step did, and it runs only on
a tag, so no rehearsal would have caught it. RP-18's conformance case reads
the two workflows as text and asserted the old `cmp`, `metadata` and
`identity` lines of crate.yml; it now asserts the lines the workflows carry
(`"$dry_crate"`), with its plants following. conformance_rp_18 passes on
this tree.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
release.yml and crate.yml: the release run starts and finds its dry-run archive; the v0.4.0 tag commit

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
Every row of the 0.4.0 register joins this line's register (472 rows);
the hand-written sources are the union of both sides. Two identifiers
collided and the release/v0.4 row keeps its number in each case: the
loop-end row of #279 is RB-22 (RB-21 on this line before the merge),
and the three SHA-256 rows of #126 stage 2 are SL-06, SL-07 and SL-08
(SL-05 to SL-07 before). The bootstrap validator and the Rust oracle
carry both the loop-end tail check of #279 and the machine callee and
overflow value of #91 stage 4; §17.13 narrows the loop-end tail to a
call. The migrate API uses the compiler's own digest, since this line
no longer ships the sha2 crate. Locks, records, goldens, packages,
fixtures and verified records regenerate in the next commits.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
The renderer record moves with the merged runtime and the language-1.2
compiler-semantics ID with it (fc363733… → bf827c3c…; 1.1 and 1.0 are
unchanged): locks, goldens, calculus fixtures, the production export,
the negative and positive fixtures, provenance, and the generated
documents are rewritten by their writers. The specification sources are
formatted, as `NS-01` requires. `SL-07` copies an example that imports a
source package beside the workspace file and its packages, as the other
cases do, since a project outside a workspace cannot declare one
(`LLC0108`); the rule reached this line with the merge.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
…ut the Atlas, on the merged identity

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
The case appended " tampered" to a register row and required the word
to be absent after the port; the merged text has "tampered" of its own
(LLR3007, EX-09), so the check failed on every host with the port right.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013QppGKetUURseXowZun9KY
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants