Skip to content

chore(release): t27c 0.5.0 -- the Verified Compute tool chain reaches the registry (Closes #7602) - #7603

Merged
gHashTag merged 1 commit into
masterfrom
claude/release-t27c-0.5.0
Oct 7, 2026
Merged

gHashTag merged 1 commit into
masterfrom
claude/release-t27c-0.5.0

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 7, 2026

Copy link
Copy Markdown
Owner

Closes #7602

Bumps t27c 0.4.0 -> 0.5.0 in the files release.yml's VERSION TRUTH gate reads for a t27c-v* tag: bootstrap/Cargo.toml and .zenodo.json, plus the t27c entry in Cargo.lock (cargo metadata --locked: consistent). No code change.

Why now: crates.io has t27c 0.4.0 (2026-09-21). Since then master gained the tool half of epic #6655. t27c silicon writes signed receipts that name the die. t27c receipt-key makes the signing key. t27c run-record judges a run's completeness, authentication, die level and independence. All of it is decided by code generated from specs/verified/*.t27, and signed and hashed with the .t27 crypto. None of it is in the registry crate.

Minor, not patch: new subcommands are additive; the seven backends are untouched.

Release order: the tag t27c-v0.5.0 goes on the master commit that carries this PR and #7594 (crypto wiring), #7598 (R3-2 citable run), #7601 (R3-3 tool half). Then release.yml runs preflight: manifests at 0.5.0, cargo publish --dry-run, and crates.io has no t27c 0.5.0 yet. Only after that does it publish.

ZENODO_DEPOSITION_T27C stays unset, so no DOI is minted, as in 0.4.0.

Checks: lefthook run pre-commit (own-language PASS, foreign-budget PASS), tri hooks pre-commit PASSED, L1 PASSED.

🤖 Generated with Claude Code

… the registry (Closes #7602)

t27c on crates.io is 0.4.0 (2026-09-21). Since then the compiler on master
gained the tool half of epic #6655: `t27c silicon` writes signed receipts
naming the die, `t27c receipt-key` makes the signing key, and
`t27c run-record` judges a run's completeness, authentication, die level
and independence -- all decided by code generated from specs/verified/*.t27,
signed and hashed with the .t27 crypto. A user who installs from the
registry can do none of it. Minor rather than patch: new subcommands,
additive; the seven backends are untouched.

Both files the release pipeline's VERSION TRUTH gate reads for a t27c-v*
tag are bumped together (bootstrap/Cargo.toml, .zenodo.json), and the
t27c entry in Cargo.lock with them (cargo metadata --locked: consistent).

ZENODO_DEPOSITION_T27C stays unset, so zenodo-publish.yml skips and mints
no DOI, as in 0.4.0.

Closes #7602

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

github-actions Bot commented Oct 7, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-07 20:43:13 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 48
PRs with All Checks Green 2
READY 1
FAILING 48
PENDING 0
NO CHECKS YET 0

These columns do not partition: 1 + 48 + 0 + 0 = 49, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=39fa3908f9fd != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@gHashTag
gHashTag merged commit 6381c73 into master Oct 7, 2026
34 of 37 checks passed
dmitrii-f-t27 added a commit that referenced this pull request Oct 8, 2026
Refs #7669, #7758, #7761

t27c 0.5.0 refuses the DNA of every die whose FUSE_DNA tail is not die A's `0x4F`. On die B, `t27c silicon` passed but wrote `device_dna: null` (#7759). The fix is #7761. R3-3 operators build from the tag, so the fix only reaches their receipts through a new tag.

This PR bumps the three files that release.yml's VERSION TRUTH gate reads for a `t27c-v*` tag, all in one commit, as #7603 did for 0.5.0:
- `bootstrap/Cargo.toml` 0.5.0 → 0.5.1
- `.zenodo.json` 0.5.0 → 0.5.1
- `Cargo.lock` t27c entry; `cargo metadata --locked` is consistent

No code change. Patch, not minor: it relaxes a rule in the verified specs, adds no subcommand, and leaves the backends untouched. `ZENODO_DEPOSITION_T27C` stays unset, so no DOI is minted.

Order: #7761 and this PR merge, then the tag `t27c-v0.5.1` goes on that master commit. After it, the ternary_link seal is re-saved by the tag build (as #7681 did) and die B is rerun for #7759.

🤖 Generated with [Claude Code](https://claude.com/claude-code)
EOF2
chore(release): t27c 0.5.1 -- device DNA is named on every die (Closes #7762)

t27c 0.5.0 refuses the DNA of any die whose FUSE_DNA tail is not die
A's 0x4F (#7758, fixed by #7761). R3-3 receipts come from the tag build,
so the fix needs a tag. Both manifests the release VERSION TRUTH gate
reads for t27c-v* (bootstrap/Cargo.toml, .zenodo.json) and the t27c
entry in Cargo.lock move together (cargo metadata --locked: consistent).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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.

release: t27c 0.5.0 -- the Verified Compute tool chain reaches the registry

2 participants