Skip to content

claw: recipes table AF4 round 1 -- two verified fpga-build rows - #9296

Merged
gHashTag merged 1 commit into
masterfrom
claw-af4-harvest
Oct 12, 2026
Merged

gHashTag merged 1 commit into
masterfrom
claw-af4-harvest

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

AF4 round 1 (#9285): harvest verified recipes into the table. Two fpga-build heals the night loop actually ran and merged this week become rows, keyed to the real stem.

The two rows

  • elab-baseline-raise (fix(fpga): record die_probe in the elab baseline -- 205 -> 206 #9212, merged a076a8e) -- the elaboration ratchet rose when die_probe landed with one Enable of unknown task error, a class the tool's docstring already documents as emitter debt (yosys/Vivado accept, iverilog refuses). Repair: check_elab_ratchet.py --update-baseline records the module at its measured count, and the six 0-error siblings get rows in the same commit. The COND is the guardrail: a NEW error class is reported, never baselined.
  • physics-move (specs: move the three real-valued models from fpga/ to physics/ #9233, merged) -- three specs were f64 math models living under specs/fpga/; yosys parse-rejects function real on principle (iverilog accepts, which is why elaboration said 0). Repair: move to specs/physics/, re-seal on the lab, re-path suite_expectations.json by line surgery, re-run the corpus ratchets. Integer-logic siblings (secded) stay where they are.

Both C_AUTO: agent-executed end to end, verified by the gates they heal, auto-merged.

Table shape

21 rows; census C_AUTO 15 / C_REVIEW 3 / C_OWNER 3; fpga-build carries 2 rows (one workflow, two checks, different repairs -- the stem join allows exactly this). Index pins extended to rows 19-20; census, gate-count, routing and instrument tests updated in the same commit (the table cannot drift silently).

Verification

Lab: typecheck OK (0 errors), 11/11 tests, 0 FAIL, 0 vacuous. ASCII clean. dupe_scan --like: no new groups (str_eq 2 already in the ledger from #9290).

Refs #9285, #9281; part of #9271. Next harvest rounds: trinity/trios/fpga CI patterns (the related-repos pass) and the lean-completeness family.

🤖 Generated with Claude Code

…-baseline-raise, physics-move)

Harvest pass #9285: the night loop ran and verified two fpga-build heals this
week; both land as rows, keyed to the real stem.

- elab-baseline-raise (#9212): iverilog error class documented as emitter debt
  ("Enable of unknown task") -> check_elab_ratchet.py --update-baseline records
  the module at its measured count plus 0-error siblings in the same commit.
  COND holds the line: a NEW error class is reported, never baselined.
- physics-move (#9233): real-typed f64 math specs under specs/fpga that yosys
  rejects at "function real" -> move to specs/physics, re-seal on the lab,
  re-path suite_expectations by line surgery. Integer-logic siblings stay.

Table 19 -> 21 rows; census C_AUTO 15 / C_REVIEW 3 / C_OWNER 3; fpga-build
carries 2 rows (one workflow, two checks, different repairs); index pins for
rows 19-20; census/gate-count/routing/tool tests updated in the same commit.

Lab: typecheck OK, 11/11 tests, 0 FAIL, 0 vacuous. No new duplicate groups.

Refs #9285, #9281. Part of #9271.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@gHashTag
gHashTag enabled auto-merge (squash) October 12, 2026 17:48
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-12 17:48:20 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 49
PRs with All Checks Green 1
READY 0
FAILING 49
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 49 + 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)=e769ef021713 != 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 90fc25b into master Oct 12, 2026
24 of 29 checks passed
gHashTag added a commit that referenced this pull request Oct 12, 2026
…(slices_equal, starts_with) (#9299)

Fresh master dbf1920 re-scan first (recipe row dupe-bless, specs/claw/recipes.t27):
exactly the two known 2x groups, nothing new -- engine.t27 (#9292), the recipes
table rounds (#9296/#9297) and the gates regen added no unledgered bodies. Port
wave quiet since #9275 (2026-10-12 16:20Z, >26 h). 712 bodies / 212 groups /
2003 specs; 199 -> 201 blessed groups, two scoped lines added, rescan clean.

Refs #9281.

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.

1 participant