ci(fpga-build): run the gate-precondition control when a gate it guards changes, and after a red ratchet (Refs #5453)

[dmitrii-f-t27]

Refs #5453

tools/check_gate_preconditions.py already catches what #5183 did to the seal gate. It never ran where it could have, and this PR makes it run. The restore itself is #5454; this PR does not touch tools/check_seal_coverage.py.

What the control says

Measured locally on master d4c4f2f, with t27c rebuilt from it:

tools/check_seal_coverage.py	python3 tools/check_gate_preconditions.py
master's one-line file	exit 1: VACUOUS check_seal_coverage.py [bare] exits 0 with nothing to check
the 687-line file from bcb32d723^ (#5454)	exit 0: OK: 10 precondition(s) across 6 gates fail loudly

Why it never said so

1. Path filters. fpga-build.yml started only on two of the six gates in that table (check_vector_data.py, check_elab_ratchet.py). 131 seals are stale: the spec changed after sealing #5183 touched tools/check_seal_coverage.py and a docs/now entry, so the workflow never ran.
2. One red step skips the rest. Since a90bdac (2026-09-22), every master run of fpga-conformance has failed at step 9, "Elaboration errors may fall, never rise", and GitHub skipped steps 10-16. Step 11 is this control (run 36883584257 is the latest). So it returned no verdict for the week before 131 seals are stale: the spec changed after sealing #5183 and the two days after.

Change

• Both path lists (pull_request, push) gain the other four gates in the table, plus the helper it plants them with: check_seal_coverage.py, check_duplicate_agreement.py, check_specs_generate.py, check_specs_parse.py and tools/_prereq.py.
• The step gets if: always(), the idiom bootstrap-tests.yml and corpus-ratchet.yml already use. It reads the built t27c and the tree, never the ratchet's verdict. If t27c did not build, it reports UNRUN, which is red and true.

Expected on this PR: step 11 goes red with VACUOUS for check_seal_coverage.py, because master still carries the one-line file. It goes green when #5454 lands. fpga-conformance is red on master anyway, from step 9. Neither is a required check; the required checks are validate, check-linked-issue and parse-ratchet.

Not changed: steps 10 and 12-16, which the same ratchet skips: its own negative control, both vector-data steps, conformance vector execution, JSON structure and power regression. Whether they should run past a red ratchet is the FPGA job's call.

Independent check of #5454

No code from #5454 is in this PR. Its file is byte-identical to bcb32d723^ (checked with cmp).

• --self-check: exit 0. All 8 end-to-end cases land on the right branch; compare passes 4 of 4 and the legend covers 8 of 8.
• Real run: exit 1. Of 1404 seals, 654 newly do not hold (545 stale, 109 gen-drift).
  • The seal gate is back: a merge her verdict refused had made it one comment line #5454's body says 649 (540 stale).
  • Between its base de2a1ac and d4c4f2f, no sealed spec and no compiler source changed. So I could not reproduce 649 from the tree, and did not chase it further.
• Re-derived without the gate, using shasum, jq and t27c seal with no Python: the same 654 name/kind pairs, plus 656 that hold, 89 with no spec on disk and 5 with no spec_path.
• Planted on the real tree:
  • One comment line appended to specs/tri/agent/faculty_board.t27 gives 655, with agent_FacultyBoard.json [stale]. t27c seal --verify agrees: spec_hash MISMATCH.
  • One zeroed gen_hash_zig gives 655, with [gen-drift].
  • Both reverted: back to 654, byte-identical output.
  • Master's one-line file on the planted tree exits 0 with zero bytes of output, and so does its --self-check.
• At bcb32d723^ itself, with t27c built there: 641 of 1397 (539 stale, 102 gen-drift), not 131.
  • 131 is 131 seals are stale: the spec changed after sealing #2547's count, which 131 seals are stale: the spec changed after sealing #5158 carried over "unchanged".
  • Since the removal, 4 seals were re-sealed (S09: the Queen's task lifecycle, measured from the supervisor that runs (Closes #4828) #5360, String == is the emitter's job: two string names compare with std.mem.eql (Closes #5162) #5444) and 17 broke.
  • The 10 newly stale seals are five twin pairs, after 97296cc, A recovered assertion that actually fails: trits_to_bits round-trip in ternary_encoding #5286, contrib/solana tri-mining and depin.prove rewards still read as the TRI emission rule #5415 and Remove 86 pointless discards from 6 specs: -86 errors, no compiler change #5195.
  • The 7 new gen-drift seals are ours. S07-S10 (235f6ab, c8afd3a, 20fe650, 50bf0ee) sealed them with a t27c built at the bff21b8 pin, not with the compiler of their own commits. All four hashes match that binary. Their gen_hash_zig matches no master compiler, either at those commits or now. They were wrong on arrival, and the gate that would have said so was empty.
• Reconciled with check_seal_currency.py: it reports 592 stale gen hashes, 68 gen_hash=none, 650 current and 94 missing. That is the same partition once its two design choices are named:
  • It does not compare spec_hash. So the 21 seals whose spec changed while all four outputs stayed byte-identical read as current there.
  • It counts the 68 none seals apart. Here 27 of them read as holding, because none == none (ci-gates §314).

Baseline: unchanged, by measurement

tools/seal_baseline.txt has 94 lines, and they are exactly the 94 seals with no spec on disk or no spec_path, one for one. None is paid, none departed and none changed class, so there is nothing to tighten (§67).

Writing the 654 into it by hand would produce the same file --update-baseline writes. That is the decision #5158 left to the owner, and #5453 asks the gate to report red until it is made. It would also file seven seals that are simply wrong as debt, when the fix for those is a re-seal with master's t27c.

🤖 Generated with Claude Code

[github-actions]

📓 NotebookLM Notebook linked to this PR

• Notebook ID: b83263109fb055dc
• URL: https://notebooklm.google.com/notebook/b83263109fb055dc

This notebook contains session context, decisions, and artifacts for this work.

[github-actions]

PR Dashboard

Generated at: 2026-10-01 23:46:27 UTC

• ci(fpga-build): run the gate-precondition control when a gate it guards changes, and after a red ratchet (Refs #5453) #5489 ci(fpga-build): run the gate-precondition control when a gate it guards changes, and after a red ratchet (Refs #5453) | Created: 2026-10-01 | Updated: 2026-10-01 | 21 failing / 33 total
• fix(ci): the NOW sync gate writes nothing to the clone it runs in (Refs #5482) #5488 fix(ci): the NOW sync gate writes nothing to the clone it runs in (Refs #5482) | Created: 2026-10-01 | Updated: 2026-10-01 | 9 failing / 32 total
• Port gHashTag/trinity:fpga/openxc7-synth/led_off_test.v (Verilog, 1 module) to specs/port/trinity/fpga/openxc7-synth/led #5487 Port gHashTag/trinity:fpga/openxc7-synth/led_off_test.v (Verilog, 1 module) to specs/port/trinity/fpga/openxc7-synth/led | Created: 2026-10-01 | Updated: 2026-10-01 | 12 failing / 30 total
• Port gHashTag/trinity:src/tri/gen_work_stealing.zig (Zig, 4 functions) to specs/port/trinity/src/tri/gen_work_stealing.t #5486 Port gHashTag/trinity:src/tri/gen_work_stealing.zig (Zig, 4 functions) to specs/port/trinity/src/tri/gen_work_stealing.t | Created: 2026-10-01 | Updated: 2026-10-01 | 10 failing / 31 total
• Port gHashTag/trinity:fpga/openxc7-synth/d6_test.v (Verilog, 1 module) to specs/port/trinity/fpga/openxc7-synth/d6_test. #5485 Port gHashTag/trinity:fpga/openxc7-synth/d6_test.v (Verilog, 1 module) to specs/port/trinity/fpga/openxc7-synth/d6_test. | Created: 2026-10-01 | Updated: 2026-10-01 | 9 failing / 31 total
• Port gHashTag/trinity:src/tri/gen_ray.zig (Zig, 3 functions) to specs/port/trinity/src/tri/gen_ray.t27 #5484 Port gHashTag/trinity:src/tri/gen_ray.zig (Zig, 3 functions) to specs/port/trinity/src/tri/gen_ray.t27 | Created: 2026-10-01 | Updated: 2026-10-01 | 8 failing / 31 total
• fix(trinity): the S01 manifest skips build.zig comments and counts the vendored t27 copies apart (Refs #3563) #5474 fix(trinity): the S01 manifest skips build.zig comments and counts the vendored t27 copies apart (Refs #3563) | Created: 2026-10-01 | Updated: 2026-10-01 | 5 failing / 36 total
• Specify contributor key management and attribution boundaries #5473 Specify contributor key management and attribution boundaries | Created: 2026-10-01 | Updated: 2026-10-01 | 5 failing / 36 total
• fix(bootstrap): make cargo test compile again -- honest server gates, E0716 (#5448) #5455 fix(bootstrap): make cargo test compile again -- honest server gates, E0716 (#5448) | Created: 2026-10-01 | Updated: 2026-10-01 | 8 failing / 45 total
• The seal gate is back: a merge her verdict refused had made it one comment line #5454 The seal gate is back: a merge her verdict refused had made it one comment line | Created: 2026-10-01 | Updated: 2026-10-01 | 5 failing / 34 total
• The Queen's accept is the merge: dispatch.t27 says so, and its harness measures it #5452 The Queen's accept is the merge: dispatch.t27 says so, and its harness measures it | Created: 2026-10-01 | Updated: 2026-10-01 | 6 failing / 32 total
• docs(tri): mark the Solana tri-mining program and depin.prove rewards superseded by mint-on-acceptance (Closes #5412) #5413 docs(tri): mark the Solana tri-mining program and depin.prove rewards superseded by mint-on-acceptance (Closes #5412) | Created: 2026-10-01 | Updated: 2026-10-01 | 6 failing / 32 total
• feat(automation): ball-grants -- the mail cards, to others only by the owner's grant (Closes #5387) #5411 feat(automation): ball-grants -- the mail cards, to others only by the owner's grant (Closes #5387) | Created: 2026-10-01 | Updated: 2026-10-01 | 5 failing / 32 total
• spec(automation): kanban-card-chat -- a card opens its agent conversation #5389 spec(automation): kanban-card-chat -- a card opens its agent conversation | Created: 2026-10-01 | Updated: 2026-10-01 | 5 failing / 32 total
• Port gHashTag/BrowserOS:trios/agent-server/apps/server/src/lib/container/image-loader.ts (TypeScript, 2 functions) to sp #5320 Port gHashTag/BrowserOS:trios/agent-server/apps/server/src/lib/container/image-loader.ts (TypeScript, 2 functions) to sp | Created: 2026-09-30 | Updated: 2026-09-30 | 8 failing / 31 total
• A "using" in prose became an import; and the per-spec delta check compared strings #5215 A "using" in prose became an import; and the per-spec delta check compared strings | Created: 2026-09-29 | Updated: 2026-09-29 | 12 failing / 41 total
• Auto-loop 2026-09-29: silent-discard parser family + loop infrastructure #5084 Auto-loop 2026-09-29: silent-discard parser family + loop infrastructure | Created: 2026-09-28 | Updated: 2026-09-29 | 10 failing / 45 total
• Carry wave-loop-898's stranded work: green lean 24/24 (fixes red lean CI), mcp registry spec, research corpus #5081 Carry wave-loop-898's stranded work: green lean 24/24 (fixes red lean CI), mcp registry spec, research corpus | Created: 2026-09-28 | Updated: 2026-09-28 | 10 failing / 36 total
• chore: drop build artifacts tracked since ea15cd54c (#1399), ignore *.rlib #5078 chore: drop build artifacts tracked since ea15cd54c (#1399), ignore *.rlib | Created: 2026-09-28 | Updated: 2026-09-28 | 7 failing / 40 total
• Port gHashTag/BrowserOS:trios/agent-server/apps/server/src/api/services/queen-brief-shape.ts (TypeScript, 2 functions) t #4911 Port gHashTag/BrowserOS:trios/agent-server/apps/server/src/api/services/queen-brief-shape.ts (TypeScript, 2 functions) t | Created: 2026-09-27 | Updated: 2026-09-27 | 11 failing / 31 total
• spec: agent-conversations — several chats per agent, as a compiler-checked contract #4892 spec: agent-conversations — several chats per agent, as a compiler-checked contract | Created: 2026-09-27 | Updated: 2026-09-27 | 10 failing / 31 total
• Seals: 0 of 1072 specs verify, 928 with an unchanged spec_hash — gate it or remove it #4887 Seals: 0 of 1072 specs verify, 928 with an unchanged spec_hash — gate it or remove it | Created: 2026-09-27 | Updated: 2026-09-27 | 12 failing / 40 total
• spec(automation): agent-conversations — several chats per agent #4840 spec(automation): agent-conversations — several chats per agent | Created: 2026-09-26 | Updated: 2026-09-26 | 9 failing / 32 total
• Wave Loop 473 — compiler-backend aggregate tail / live cold-POR CCLK sweep #4801 Wave Loop 473 — compiler-backend aggregate tail / live cold-POR CCLK sweep | Created: 2026-09-24 | Updated: 2026-09-24 | 11 failing / 31 total
• Bare for statements: 2 loops recovered, and 235 of the 240 sites I cited were out of scope #4777 Bare for statements: 2 loops recovered, and 235 of the 240 sites I cited were out of scope | Created: 2026-09-24 | Updated: 2026-09-24 | 10 failing / 31 total
• Generic type parameters dropped from the declaration and kept in the body: -23 errors #4745 Generic type parameters dropped from the declaration and kept in the body: -23 errors | Created: 2026-09-24 | Updated: 2026-09-24 | 12 failing / 31 total
• Implement the 8 empty function bodies in specs/vsa/packed_vsa.t27 #4729 Implement the 8 empty function bodies in specs/vsa/packed_vsa.t27 | Created: 2026-09-24 | Updated: 2026-09-24 | 12 failing / 31 total
• Wave 684: occlusion four deep, and the two files no list ever reached #4660 Wave 684: occlusion four deep, and the two files no list ever reached | Created: 2026-09-23 | Updated: 2026-09-23 | 10 failing / 31 total
• Capture given/when/then and render them: 6540 clauses, and 169 specs share one unfilled placeholder #4642 Capture given/when/then and render them: 6540 clauses, and 169 specs share one unfilled placeholder | Created: 2026-09-23 | Updated: 2026-09-23 | 10 failing / 31 total
• Implement part 2 of 2 of the empty bodies in specs/git/diff.t27 (7 functions) #4634 Implement part 2 of 2 of the empty bodies in specs/git/diff.t27 (7 functions) | Created: 2026-09-23 | Updated: 2026-09-23 | 10 failing / 31 total
• feat(specs/skills): trinity/connect-accounts #4600 feat(specs/skills): trinity/connect-accounts | Created: 2026-09-22 | Updated: 2026-09-22 | 7 failing / 32 total
• Port part 1 of 2 of tools/check_withdrawn_live.py to specs/port/tools/check_withdrawn_live.t27 (8 functions) #4581 Port part 1 of 2 of tools/check_withdrawn_live.py to specs/port/tools/check_withdrawn_live.t27 (8 functions) | Created: 2026-09-22 | Updated: 2026-09-22 | 9 failing / 31 total
• Port part 1 of 2 of tools/check_coq_in_build.py to specs/port/tools/check_coq_in_build.t27 (8 functions) #4578 Port part 1 of 2 of tools/check_coq_in_build.py to specs/port/tools/check_coq_in_build.t27 (8 functions) | Created: 2026-09-22 | Updated: 2026-09-22 | 8 failing / 31 total
• Implement the 1 empty function body in specs/test_framework/verilog_bench_harness.t27 #4565 Implement the 1 empty function body in specs/test_framework/verilog_bench_harness.t27 | Created: 2026-09-21 | Updated: 2026-09-21 | 0 failing / 2 total
• Implement part 2 of 2 of the empty bodies in specs/vm/jit_semantics.t27 (4 functions) #4563 Implement part 2 of 2 of the empty bodies in specs/vm/jit_semantics.t27 (4 functions) | Created: 2026-09-21 | Updated: 2026-09-21 | 8 failing / 31 total
• The C tests handed cc a clang-only flag: 23 red on gcc, and the silent half compiled nothing #4551 The C tests handed cc a clang-only flag: 23 red on gcc, and the silent half compiled nothing | Created: 2026-09-21 | Updated: 2026-09-22 | 5 failing / 44 total
• classify: the browser could not tell a Markdown file from a broken spec (refs #2133) #4541 classify: the browser could not tell a Markdown file from a broken spec (refs #2133) | Created: 2026-09-21 | Updated: 2026-09-22 | 7 failing / 41 total
• Implement the 1 empty function body in specs/test_framework/graph_drift_detection.t27 #4536 Implement the 1 empty function body in specs/test_framework/graph_drift_detection.t27 | Created: 2026-09-21 | Updated: 2026-09-21 | 8 failing / 31 total
• Nothing builds bindings/wasm-explorer, which is how the artifact it replaces drifted in the first place #4498 Nothing builds bindings/wasm-explorer, which is how the artifact it replaces drifted in the first place | Created: 2026-09-21 | Updated: 2026-09-21 | 6 failing / 30 total
• specs/functions: render-job-run, morph-images-generate -- a failed delivery is not delivered #4123 specs/functions: render-job-run, morph-images-generate -- a failed delivery is not delivered | Created: 2026-09-17 | Updated: 2026-09-17 | 4 failing / 30 total
• specs/functions: payment-ai-server-process -- claim-invoice before crediting #4116 specs/functions: payment-ai-server-process -- claim-invoice before crediting | Created: 2026-09-17 | Updated: 2026-09-17 | 4 failing / 30 total
• specs/functions: crm-proactive-sweep -- one step per seller after the 502 runs #4113 specs/functions: crm-proactive-sweep -- one step per seller after the 502 runs | Created: 2026-09-17 | Updated: 2026-09-17 | 4 failing / 30 total
• specs/functions: withdraw five functions from the registry (code-only/unregistered) #4038 specs/functions: withdraw five functions from the registry (code-only/unregistered) | Created: 2026-09-17 | Updated: 2026-09-17 | 4 failing / 30 total
• Feed the swarm from CI, with fuel that does not run out #3978 Feed the swarm from CI, with fuel that does not run out | Created: 2026-09-17 | Updated: 2026-09-17 | 6 failing / 28 total
• specs: bind every tool and skill to the Queen, and name a domain lead #3833 specs: bind every tool and skill to the Queen, and name a domain lead | Created: 2026-09-15 | Updated: 2026-09-15 | 4 failing / 30 total
• agent-provider-chain: probe 4 of the reserve, 402 with max_tokens bounded #3683 agent-provider-chain: probe 4 of the reserve, 402 with max_tokens bounded | Created: 2026-09-15 | Updated: 2026-09-15 | 0 failing / 30 total
• specs/automation: inngest-functions-status.t27 (lastRun без проб, lastProbe отдельно) + inngest-probe-suite VERSION 2 #3593 specs/automation: inngest-functions-status.t27 (lastRun без проб, lastProbe отдельно) + inngest-probe-suite VERSION 2 | Created: 2026-09-13 | Updated: 2026-09-13 | 5 failing / 30 total
• fix(ci): make the JSON gate strict, and make the one file it refuses legal #3395 fix(ci): make the JSON gate strict, and make the one file it refuses legal | Created: 2026-09-07 | Updated: 2026-09-07 | 0 failing / 30 total
• coq(CorePhi): replace proofs that were never checked by a compiler #3343 coq(CorePhi): replace proofs that were never checked by a compiler | Created: 2026-09-06 | Updated: 2026-09-06 | 5 failing / 32 total
• chore(ci)(deps): bump actions/attest-build-provenance from 1 to 4 #3001 chore(ci)(deps): bump actions/attest-build-provenance from 1 to 4 | Created: 2026-09-01 | Updated: 2026-09-14 | 1 failing / 31 total

Summary

Status	Count
Total Open PRs	50
PRs with Failing Checks	47
PRs with All Checks Green	3
READY	3
FAILING	47
PENDING	0
NO CHECKS YET	0

Seal Status

• ⚠️ STALE -- sha256(compiler.rs)=b7d5cc5c4cf1 != 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).