chore(rivet): RQ-60-CANARY + RQ-60-A64IMPORT implemented — the 4th and 5th orphaned flips, plus a floor left slack - #1074
Merged
Merged
Conversation
…d 5th orphaned flips RQ-60-FLIPCOUPLE is now demonstrating itself. That artifact was filed BECAUSE three status flips were orphaned in v0.59; here are two more, in the release that scopes the fix: RQ-60-CANARY its work SHIPPED IN v0.59.0 (#1061, tag-verified) — the artifact was written describing already-delivered work and landed as `proposed` anyway RQ-60-A64IMPORT #1071 merged minutes ago and did not flip its own status Verified ON MAIN rather than inferred from the merge list: expansion_canary_gate_1021.py present, 1 CI invocation, `git tag --contains` reports v0.59.0 aarch64_import_dispatch_1017_differential.py present, 1 CI invocation ALSO TIGHTENED THE FLOOR 472 -> 473. #1071 added SWVER-025 without raising it. A floor only fails BELOW itself, so 473-measured against 472-pinned would have passed QUIETLY — slack, not red. The gate's own rule is that a PR adding artifacts raises the floor in that PR, and leaving it loose is the first crack in the discipline the gate exists to create. That is the same failure shape twice in one commit: a number that drifts in the direction that still looks green. Both are exactly what RQ-60-FLIPCOUPLE is scoped to make mechanical instead of vigilant. Verified BY ID: both resolve `[implemented]`; CI-filter OURS=0; measured 473 = floor 473; claim_check exit 0. Refs #1064, Refs #1017, Refs #1021 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01YJK5LZZEkV5smCY1jKn18L
#1073 merged as e6a3b27 and did not flip its own status. That is three in v0.59 and now three in v0.60, in the release whose RQ-60-FLIPCOUPLE artifact exists to fix exactly this. Verified ON MAIN, not inferred: 8 `__aeabi_ul2f`/`__aeabi_f2lz` references in crates/synth-synthesis/src/, and scripts/repro/aeabi_i64_float_1069_differential.py present. Six for six, every one caught by querying AFTER the merge rather than by any mechanism. That is the argument for RQ-60-FLIPCOUPLE stated as data rather than as a worry: a discipline that depends on somebody remembering fails on the tick they do not. Verified BY ID: CANARY / A64IMPORT / VFPPRESSURE all `[implemented]`; CI-filter OURS=0; measured 473 = floor 473; claim_check exit 0. Refs #1064, Refs #1069 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01YJK5LZZEkV5smCY1jKn18L
avrabe
force-pushed
the
chore-v060-flips
branch
from
August 26, 2026 22:20
6b813bc to
f4fd282
Compare
I flipped it mechanically on "#1073 merged" without reading the artifact's own DEFINITION OF DONE: those three symbols reaching `nm -> T` — 5 of 5 stages exported Increment 1 did not achieve that. It cleared `position#tick` on single-precision cortex-m7; on **m7dp — jess's actual flight core — all three symbols still fail on main**. Increment 2 (#1075, open) is what delivers 5/5. The flip belongs there. This is the MIRROR of the orphaned-flip class I have been chasing all release. Six times I caught a status left `proposed` over shipped code; here I set one to `implemented` over code that had not shipped the thing the artifact promises. Both directions under-report the truth, and RQ-60-FLIPCOUPLE should derive status from EVIDENCE precisely so neither is a judgement call — "the PR merged" is not the same claim as "the artifact's stated outcome holds". ALSO: increment 2's written plan is marked RETIRED BY MEASUREMENT rather than deleted, so it is not re-proposed. The plan was "extend the S pool into callee-saved S16..S31". A discriminating variant — same pressure, different RESIDENCE — showed the wall is not capacity: 60 simultaneously-live f32 on the OPERAND STACK compile while 14 in HOMED LOCALS do not. Pool extension moves the wall 13 -> ~29: a bigger number and the same wall further out, paid with prologue/epilogue push/pop. Recorded because a retired plan that leaves no trace gets proposed again. v0.60 reads 2/8; VFPPRESSURE lands with #1075. CI-filter OURS=0, measured 473 = floor 473, claim_check exit 0. Refs #1069, Refs #1064 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01YJK5LZZEkV5smCY1jKn18L
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
RQ-60-FLIPCOUPLEis demonstrating itself. That artifact exists because three status flips were orphaned in v0.59. Here are two more — in the release that scopes the fix.The two flips
RQ-60-CANARYproposedanywayRQ-60-A64IMPORTVerified on main, not inferred from the merge list:
And a floor left slack
ARTIFACT_FLOOR472 → 473. #1071 addedSWVER-025without raising it.A floor only fails below itself — so
473measured against472pinned would have passed quietly. Not red, just loose. The gate's own rule is that a PR adding artifacts raises the floor in that PR, and leaving it slack is the first crack in the discipline the gate exists to create.That's the same failure shape twice in one commit: a number drifting in the direction that still looks green. A stale
proposedunder-reports scope; a slack floor under-detects a skipped file. Neither goes red, and both are exactly whatRQ-60-FLIPCOUPLEis scoped to make mechanical rather than dependent on somebody noticing.Verification
Validated by id, per #1064 — not by absence of an error: