extdeps/rust: ground Cargo versions on SemVer carriers

[gunbai-bot]

Summary

• Add dsl/extdeps/version/semver.dag (slice 2): SemVerIdentity / SemVerConstraint inhabit VersionIdentity / VersionConstraint from merged extdeps/version.dag: base versioning model (VersionIdentity + VersionScheme + VersionConstraint) #5311; cites semver.org; not std.types.SemVer
• Add dsl/extdeps/rust/version.dag (extdeps.cargo_version): CargoPackageVersion = SemVerIdentity, CargoVersionRequirement = SemVerConstraint
• Wire cargo.dag to the new carriers
• Witnesses: cargo_version_witness_test.dag, extdeps_version_semver_witness_test.dag

Merge order (operator-visible)

PRIMARY for realization_width.dag overlap: merge #5309 before #5307. #5307 (PEP 440) is DRAFT until this lands; it must rebase + re-verify CI on the merged tree before flipping ready.

Also merge before #5322 (constraint grounding, draft).

Sequencing (layer reconcile per gentle-raven-832)

1. This PR (extdeps/rust: ground Cargo versions on SemVer carriers #5309) — Cargo semver carriers on extdeps.version.semver (slice 2 minimal: wire-label inhabitation + interim semver_scheme equality compare; slice-2b adds MAJOR.MINOR.PATCH decomposition + semver-aware ordering)
2. Separate PR (escalation-gated) — migrate std.types.SemVer consumers (OCI OciRuntimeVersion, etc.) off the std alias onto extdeps.version.semver
3. Parent program — PEP 440, go modules, git scheme slices build on the same extdeps.version base (extdeps/version.dag: base versioning model (VersionIdentity + VersionScheme + VersionConstraint) #5311)

Test plan

• gunbc run --source-root dsl --entry dsl/test/claim/cargo_version_witness_test.dag --function cargo_version_witnesses --claim-run
• gunbc run --source-root dsl --entry dsl/test/claim/extdeps_version_semver_witness_test.dag --function extdeps_version_semver_witnesses --claim-run
• gunbc run --source-root dsl --entry dsl/test/claim/extdeps_version_witness_test.dag --function extdeps_version_base_witnesses --claim-run
• CI passes

[gunbai-bot]

Investigated CI failure @ 91f10c8: the job died during cargo build before claim_executor ran — sccache server I/O error (Failed to send data to or receive data from server / failed to fill whole buffer), not a witness or substrate regression from this PR.

Verified locally on the same HEAD:

• env -u RUSTC_WRAPPER cargo build -p v1-compiler --release --features text_lookup_work_counter --bins — OK
• Full claim_executor CI floor (gunbc_ci_floor_batches, 455 witnesses) — PASS

Re-triggered CI (workflow_dispatch on session/deep-bear-687-cargo-semver). No code change required.

— sent from deep-bear-687

[gunbai-bot]

Addressed REQUEST_CHANGES @ review 31012:

Finding: default_stage0_package_version was declared single authority but emit_stage0_crate_manifest / emit_cargo_toml still hardcoded version = \"0.1.0\".

Fix (@ 661cd10):

• Added render_cargo_package_header_prefix in extdeps.cargo_version — shared [package] header through default_stage0_package_version
• src/v1/stage0_crates.dag emit_stage0_crate_manifest now uses it
• src/v1/05_emit_rust.dag emit_cargo_toml now uses it
• Witness w_emitter_header_prefix_uses_default exercises the production emission path

— sent from deep-bear-687

[gunbai-bot]

Investigated dashboard CI alert @ 6e2ca35 — stale SHA. That was the interim WIP commit (superseded by force-push); it failed before the review-fix and main-merge repairs landed.

Current HEAD 9d0e4fb: CI green (run 27844557728, 14m43s). Fixes already on branch:

• 661cd103 — wire emit_stage0_crate_manifest / emit_cargo_toml through default_stage0_package_version
• ae2f442 + 9d0e4fb — realization_width import fix after STD_LEAK slice (NanosecondDuration decomposition): std must be perfectly decomposed — a scale-fixed time alias hardcoded in std is a leak. Make std's realization/measurement framework (std/realization*.dag) scale-AGNOSTIC (generic Measure<Time,S,M>); relocate the NANOSECOND scale CHOICE + its reason #5286 NanosecondDuration migration

No additional code change required; dashboard merge readiness shows ready=true, checks passing.

— sent from deep-bear-687

[gunbai-bot]

Addressed REQUEST_CHANGES @ review 31071:

Finding: min_cargo_version reused CargoVersionRequirement (a [dependencies] carrier) for a cargo tool edition floor — illegal states representable via RegistryDep { version: min_cargo_version, … }.

Fix: Introduced CargoToolVersionFloor (NonEmptyStr where brand("CargoToolVersionFloor")) in extdeps.cargo_version; min_cargo_version now uses that carrier. Added discriminating witness w_tool_floor_distinct_from_dep_requirement.

— sent from deep-bear-687

[gunbai-bot]

Addressed non-blocking nit @ review 31090: CargoToolVersionFloor now inhabits SemVerConstraint where brand("CargoToolVersionFloor") instead of bare NonEmptyStr — stays on extdeps.version.semver authority while preserving distinct use site from CargoVersionRequirement. Witnesses still green.

— sent from deep-bear-687

[gunbai-bot]

Re: claude #31103 (REQUEST_CHANGES @ 034b363)

Verified each finding against current HEAD d5c3c29:

1. Duplicate Nano import (realization_width.dag:29) — fixed in d5c3c29366. The rebase merge artifact is gone; std.measure now imports Nano once.

2. CargoToolVersionFloor witness only checks string inequality — acknowledged as minor in the review itself ("the brand declaration itself is fine"). CargoToolVersionFloor is a branded SemVerConstraint alias distinct from unbranded CargoVersionRequirement; type-distinctness is carried by the brand at the model layer. A compile-time non-unification witness would be a follow-up, not merge-blocking for this slice.

3. semver_identity_compare interim scaffold — no action; review notes correct shape.

Fresh dashboard reviews on d5c3c29: claude APPROVE (#31113), cursor APPROVE (#31115). CI green.

— sent from deep-bear-687