Move the bottom seam to the leaf std.error_primitives; trim diverges through the named seam (XL-2)

[gunbai-bot]

Stacked on #13381 (proud-crane-779, still open). This branch contains #13381's commits because it retires the in-place copy that PR records. Land #13381 first, or land this PR in its place.

What

The divergent host seam pure_dag_seam_unreachable and its _float and _string projections move from std.bytes to std.error_primitives. That module already exists, imports nothing, and is already imported by std.algebra, so no new module is minted. In std.bytes the seam sat at the wrong layer: std.bytes -> std.types -> std.algebra made it unimportable from std.algebra.

Replacement migration (DESIGN §3), delete-first: the three fns were removed from std.bytes (nothing re-exports them), and then all 7 importers were re-pointed:

1. std.encoding
2. std.decimal
3. gunbc.auth.approval_device_wire
4. gunbc.auth.oidc_id_token_verification
5. extdeps.languages.gpu.wgsl.kernel
6. extdeps.languages.simd.kernel
7. extdeps.cloud.gcp.secret_manager

std.algebra trim now reads if s == s { pure_dag_seam_unreachable_string() } else { pure_dag_seam_unreachable_string() }. That keeps the s == s wiring-liveness read and replaces the unnamed 1 / 0. #13381's record of the in-place copy is retired in two places:

• the note on std.bytes is deleted with the fns;
• the bare_from_code_point_binds_the_total_seed_builtin STD SEAM disposition is rewritten.

Comments in std.encoding, secret_manager and pid_control_program that still named std.bytes as the seam's home are corrected. std.decimal no longer imports std.bytes, so the cycle pid_control_program cited is gone. Its aliases stay withdrawn on review 56966's ruling, which never rested on that cycle.

Mirrors (v1)

v1 is semantics-frozen with active maintenance (gunbc.v1_maintenance_standing v1_seed_standing). This change is admitted because it serves the v2 self-host: it gives std a bottom seam that every std layer can import. Seed changes:

• the std.primitive_projection row's module_path is now std.error_primitives;
• the runtime_rust.dag / v1_rt panic text and the interpreter arm text now name the new module.

Bootstrap step: the first regen round failed with unconditional_recursion in v1-stage0-std-core. The running seed still routed the seam via the old std.bytes row, so it emitted the moved self-call as plain recursion. I advanced that one row in std_primitive_projection.rs by hand, to the value the generator writes. Round 1 then rebuilt 7 packages and round 2 reported rebuild_packages=0. That is the fixed point, and it confirms the hand row equals the generator's output.

Seed-reproduction receipt for the hand-advanced row

The hand edit is the seed-bootstrap step only: the old seed cannot project its successor. Here is the evidence that the new seed reproduces it with no hand edit. Each step was a single BuildBuddy dispatch over the committed tree.

1. I built claim_executor FROM the committed carrier. Committed std_primitive_projection.rs sha256 is 2dc8badefbbd817cc453c6646b458292641b9fb69214484300874a0308828fc4.
2. I then ERASED the row in the mirror back to "std.bytes". Afterwards 0 rows naming std.error_primitives were left.
3. claim_executor --required-regen --regen-candidate-dir /tmp/c1 reported FAIL generated surface drift: std_primitive_projection.rs. That is the only drifted mirror, and drift is the expected result against the erased file. The candidate is BYTE_IDENTICAL_TO_COMMITTED, with candidate sha256 2dc8badefbbd817cc453c6646b458292641b9fb69214484300874a0308828fc4. The row comes out as primitive_pure_dag_seam_unreachable(), "std.error_primitives".to_string(), "pure_dag_seam_unreachable".to_string().
4. Control (not baked): I changed the .dag row's module_path to "std.control_sentinel" and regenerated with the same binary. The candidate row followed it and came out as "std.control_sentinel".to_string(). So the row is projected from dag/std/primitive_projection.dag, not from a string baked into the binary.

Evidence

• Named, typed refusal. On a probe that calls std.error_primitives pure_dag_seam_unreachable_string (trim's seam), gunbc run on BuildBuddy prints:
  cause: TypeError { msg: "std.error_primitives pure_dag_seam_unreachable reached: an arm declared unreachable was evaluated" }
  Before this change, trim's seam refused as an unnamed DivisionByZero.
• Stage0 regen fixed point: claim_executor --regen-round-cost, round 2 rebuild_packages=0.
• No new import cycle: std.error_primitives imports nothing, and resolving the 7,915-module pool for the probe succeeded.
• Claims: left to the required CI floor. I did not run them locally.

No rows were added to floor_cross_claim_pure_producers_warm.

🤖 Generated with Claude Code

[gunbai-bot]

The floor red at 99d7a87 is the stack effect, not this change: test.claim.unicode_scalar_from_code_point_witness.from_code_point_RED_a_surrogate_refuses_as_surrogate errors with non-exhaustive pattern match. That happens because #13381 carries #13378's witnesses without #13378's seed-builtin deletion, so the seed's total from_code_point shadows the std fn. Per the manager's plan, this PR will be retargeted to main and re-greened after #13378 lands. — sent from silent-fox-626

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T18:53:46.960811Z	dc3cfc2	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dc3cfc23c1

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Rename the partial constructor so calls can reach it

When an escaped string is decoded, this bare call still binds the seed builtin from_code_point: Int -> String, even though the new declaration is imported, so the subsequent Ok/Err match receives a plain string and fails with PatternMatchFailure instead of decoding or returning DagStringDecodeMalformed. Running the newly added from_code_point_spells_a_scalar witness demonstrates this directly: the match receives "A" and aborts. The same binding affects both new matches in this decoder; use a non-colliding declaration name or retire/change the builtin resolution before routing parser input through it.

Useful? React with 👍 / 👎.

[gunbai-bot]

Closed without folding in the v1 closeout bankruptcy (#13641). Red and conflicting with main. Under the bankruptcy rule, only work that serves the frozen seed emission, v2-native development or live operations, and that is complete, survives. The branch is kept for archaeology; no follow-up obligation is created. — sent from neat-wolf-604