Repository navigation
Conversation
…classes first Theme: gale's requests rank highest, and the structure that keeps producing its silent classes is changed rather than patched again. Each wrong-code lane is delivered as an execution-verified fix against the reference runtime, or as a loud refusal where the input is out of scope. The two refactor themes are done when the code they replace is deleted. Twenty-one artifacts under artifacts/release-v0.91/, all `status: proposed`, `issue-scope: closes` (the only value R12 allows beside a non-claiming status; it authorises nothing until a record claims), each with a refuting command RUN at planning on 2945628 (the v0.90.0 tag commit) with a fresh 0.90.0 debug build, its output in the record. Every refuting command came back LIVE; none is refuted. RQ-91-ARCHMODEL12 is a filing, not a defect. - must, gale (a), wrong code: #1672 (RQ-91-NPADATAPTR), the #1568 residual (RQ-91-NPARESID1568), the #1604 residual (RQ-91-NPARESID1604) - must, theme: native-pointer address model (RQ-91-NPAMODEL): MUST because it is the vehicle for gale's first-ranked classes; measured from 15 shipped `fn npa_*` definitions, 3 of them refusal checks - must, gale (b): #1145's MPU-enforcement half (RQ-91-MPUENFORCE) - must, gale: #1661 (RQ-91-NPASWBOUNDS), #1191 (RQ-91-COMPEXPORTS), #1301 (RQ-91-ARMV6M) - must, theme: one image builder (RQ-91-IMAGEBUILDER), measured from 10 distinct builders in synth-cli: MUST because it carries two silent wrong-code defects that are UNFILED by decision and stated in records, RQ-91-R5DATAINIT (self-contained cortex-r5 drops data segments and global initialisers) and RQ-91-M3SELFBL (--target cortex-m3 --func-name ships a bl to itself) - should, theme: derived emulation floor (RQ-91-EMUFLOOR): merge friction and a hand-kept second copy, not wrong code; 9 of the 16 commits in c20d37b..2945628 touch ci.yml and all 9 move the floor line - should, theme: backend support levels (RQ-91-SUPPORTLEVELS): it makes the public claims honest and changes no emitted code; x86-64 enters at L3 as a maintainer-relayed request with no issue - should, carried: the spec ratchet's blindness to a corrupted .wast expectation (RQ-91-SPECDRIFT), #1673 (RQ-91-MEMKFLOAT), the acceptance corpus, still unmounted (RQ-91-CORPUS), #1460 re-measured (RQ-91-NPM10), #1136 filed (RQ-91-ARCHMODEL12), the tripwire drift check that is UNAVAILABLE on the runner (RQ-91-TRIPWIREPROT), the tripwire's blindness to an expression runs-on on 5 of 10 required jobs (RQ-91-TRIPWIREEXPR), extract_r1_numbers' noise (RQ-91-EXTRACTR1) Every record carries an `issue:` field, as R5 requires since v0.60. Where no issue was filed, the record cites the CLOSED issue where the matter was analysed (#739, #910, #1286, #1575, #1636, #1640, #1643, #1645), so its `issue-scope: closes` authorises nothing; issue_closure_check attributes the v0.90 closures of #1636 and #1643 to v0.90, not to v0.91. The release file is comments-only and names no issue. The statistics scan reports 0 digit hits and 0 word hits. The anchor moves to v0.90.0: ANCHOR_TAG and ANCHOR_PROGRAMME, at both sites each; ANCHOR_DELIVERY does not move. Read off the gate's own anchor line after setting the tag: 715 artifacts at v0.90.0 against 696 at v0.89.2, a gap of 19, equal to the 19 RQ-90 ids v0.90 added, and 134 delivery commits at both tags. Paired control: 714 and 716 fail A2, 133 and 135 fail A1, 696 fails A2 on this tree, and 715 with the tag back at v0.89.2 fails A2 and CENSUS. Refs #1672 #1568 #1604 #1145 #1661 #1191 #1301 #739 #1645 #1640 #910 #1286 #1636 #1673 #1643 #1460 #1136 #1575 Co-Authored-By: Claude Opus 5.5 <[email protected]> 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
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.
Refs #1672 #1568 #1604 #1145 #1661 #1191 #1301 #739 #1645 #1640 #910 #1286 #1636 #1673 #1643 #1460 #1136 #1575
Opens the v0.91 window: the anchor moves to v0.90.0, and
artifacts/release-v0.91/holds the scope. Modelled on the v0.90 planning PR (#1652). Nothing is delivered here. Every record isstatus: proposed, soissue_closure_check --release v0.91authorises nothing.Order of work
gale's requests rank highest, then wrong code, then the themes, then the carried items. Every refuting command was RUN on 2945628 (the v0.90.0 tag commit) with a fresh 0.90.0 debug build, and its output is in the record. Every one came back LIVE. RQ-91-ARCHMODEL12 is a filing, not a defect.
tbl()faults on an unmapped read, wasmtime returns 0x88776655; theanchorcontrol is equali32.add/select, 1 of 4 equalfn npa_*definitions, 3 of them refusal checks at 3 call sites; the 5 wired NPA oracles each exit 0 (the baseline). The grep uses[[:space:]]:git grep -Ehas no\s, and with\sthe same command prints a false 2--embedder-mpugives rc 0, an object byte-identical to no--safety-bounds, with no MPU programmingwac-composed toy; the import half does not reproduce there, and gale-nano was not availablecortex-m0plusandcortex-m0are unknown targets.textonly, with 0 segment bytes and no global init; the cortex-m3 control places bothb2: f7ff fffe bl b2i32.wast:37gives RESULT PASS, rc 0,state-skew=1E_STAGE_REQUIREDlive required-context list UNAVAILABLEruns-on090×21 and0892×12; 3 thousands-separator pairs splitWhy each theme has the priority it has.
The
issue:field. R5 requires one on every release record since v0.60. Where nothing was filed, the record cites the CLOSED issue where the matter was analysed (the #345 convention). An authorised issue that is already closed satisfies the closure obligation, soissue-scope: closesauthorises nothing there.issue_closure_checkstill attributes #1636's and #1643's closures to v0.90.Anchor
ANCHOR_TAGandANCHOR_PROGRAMMEmove at both sites each.ANCHOR_DELIVERYdoes not move. The values are read off the gate's own line:That is 715 against 696 at v0.89.2, a gap of 19, equal to the 19 RQ-90 ids. Paired control:
release-v0.91/directory, the gate fails A0 and CENSUS.Gates, on this branch
--release v0.91: 0 digit hits and 0 word hits. The release file is comments-only and names no issue.release under cut: v0.91,steps-1-2 filing VISIBLE: RQ-91-ARCHMODEL12, failures 0.--release v0.91 --since-tag v0.90.0 --allow-unclosed: 0 authorised and 0 held open. It reports VACUOUS AUTHORISED SET, which is the planning state, and rc 0 with--allow-vacuous-authorised. The v0.90 audit is unchanged: 15 authorised, 2 held open, 0 failures.run:steps, each run whole underbash -e -o pipefail: 0 non-zero.pinned required contexts match branch protection, rc 0.Squash simulation
The squash subject is pinned to this PR's title,
v0.91 planning: anchor at v0.90.0 and the v0.91 scope, gale's silent classes first (#1675). In a scratch clone, that subject was committed over 2945628 with this PR's head tree. The tree is identical to head 4f13961.status_evidence_checkexits 0 with 0 failures. The anchor readsv0.90.0 — 134 … 715 …, lag 0, and the window reads0 delivery-shaped commits since v0.90.0. The planning subject is not delivery-shaped.v0.91 NOSUCHLANE (#1672): native-pointer provenance decided once (#1675)exits 1,FAIL R10-lane: … names artifact RQ-91-NOSUCHLANE, which no release file defines.Not to be merged by this lane.
🤖 Generated with Claude Code
https://claude.ai/code/session_01YJK5LZZEkV5smCY1jKn18L