diff --git a/CHANGELOG.adoc b/CHANGELOG.adoc index 7b42281..51104ca 100644 --- a/CHANGELOG.adoc +++ b/CHANGELOG.adoc @@ -1,5 +1,21 @@ == Changelog +== 2026-11-19 — Proof receipt at 201e05b, the merge of #58; 39 rows are PWG-2ii + +* `verification/receipts/resource-algebra-proofs_praxis.deed` now describes + source commit `201e05b` (the merge of #58), from `proofs.yml` run + `44435465068` on main: Agda job `135263084676`, Rocq `135263084711`, Lean + `135263084426`. The `Proofs verdict` job and the coverage-exemption job + passed in the same run. Each ledger row was printed by + `scripts/proof-receipt.sh --run-row`. +* Census: of 40 theory rows, 39 are PWG-2ii and 1 (M-2) is ungraded. P-1 to + P-4 and K-1 to K-4 have their first grade. No row is PWG-1: only the Rocq + job runs in an image that ships its checker (PN-7). +* This is the first receipt of the Lean leg and the M-2 bridge under `lake`. +* The three runs at `15fa113` leave the ledger and are listed under "Earlier + receipts" in `PROOF-STATUS.adoc`. The `receipt_run_scope` hook that + limited them to the rows they checked is removed with them. + == 2026-11-19 — Parallel composition and the loop grade (P-1 to P-4, K-1 to K-4); every receipt superseded (40 rows ungraded for now) * `src/ResourceAlgebra/Parallel.agda` adds the `IsParallel` layer: a diff --git a/docs/status/PROOF-NEEDS.adoc b/docs/status/PROOF-NEEDS.adoc index 2aad378..6dcf9b1 100644 --- a/docs/status/PROOF-NEEDS.adoc +++ b/docs/status/PROOF-NEEDS.adoc @@ -81,7 +81,7 @@ derivation, 2026-11-17) in Rocq and in Lean as PN-2 and PN-3 do, and record their statement correspondence. *Done* in resource-algebra#45: both legs re-prove the three rows by hand, axiom-free, and the statement correspondence is recorded in `PROOF-STATUS.adoc`. The receipt at -`15fa113` records that run; the rows' current grades are in the Census of +`15fa113` recorded that run; the rows' current grades are in the Census of `PROOF-STATUS.adoc`, derived from the receipt. | `verification/proofs/coq/`, `verification/proofs/lean4/` @@ -95,13 +95,13 @@ and Lean jobs run in base images pinned by digest (`debian:trixie-slim`, install their checker at run time (`agda-bin 2.6.4.3-1+b2` by apt, the Lean 4.35.0 archive against a SHA256), which the ruling below does not count as isolated. Until this closes no row can be PWG-1: the receipt at -`15fa113` names an image for the Rocq leg only. *OPEN, blocked* on the +`201e05b` names an image for the Rocq leg only. *OPEN, blocked* on the `prover-images` repository, which does not exist yet. See "PN-7 ruling and route" below. Note: the receipt at `47de4a8` recorded PWG-1 for 27 rows; it was graded before standards#1625 (2026-11-16) made `:image` necessary for PWG-1. The -current receipt, at `15fa113`, and the Census in `PROOF-STATUS.adoc` are +current receipt, at `201e05b`, and the Census in `PROOF-STATUS.adoc` are derived by the current reader. | `.github/workflows/proofs.yml` (`proofs`, `lean-leg`); `verification/receipts/` diff --git a/docs/status/PROOF-STATUS.adoc b/docs/status/PROOF-STATUS.adoc index f63c1f2..ee3427a 100644 --- a/docs/status/PROOF-STATUS.adoc +++ b/docs/status/PROOF-STATUS.adoc @@ -16,33 +16,20 @@ asserted. The values, weakest first, are `ungraded`, `3`, `2i`, `2ii` and == Census -Out of 40 theory rows: PWG-1: 0 · PWG-2ii: 0 · PWG-2i: 0 · PWG-3: 0 · -ungraded: 40. No leg has a current receipt, so nothing is graded (§8.4). -The proof receipt -(`verification/receipts/resource-algebra-proofs_praxis.deed`) names source -commit `c31bf1a9ae9d186674c7ed79c6c9b018a1d59c2b`, the commit on main that -the change below was made on. The +Out of 40 theory rows: PWG-1: 0 · PWG-2ii: 39 · PWG-2i: 0 · PWG-3: 0 · +ungraded: 1 (M-2). The grades hold at source commit +`201e05bf3ab7acaa4a7a4a2d24f87c561d4e7f23` only, the merge of +resource-algebra#58 and the commit the proof receipt describes +(`verification/receipts/resource-algebra-proofs_praxis.deed`). The standards reader (`deed_lint.js --proof-receipt-json`), as pinned by -`scripts/proof-receipt.sh` (standards `f4e83044`), derives every grade +`scripts/proof-receipt.sh` (standards `f4e83044`), derives every grade there from the legs and receipts, and refuses a recorded grade that differs (§8.4). -Why no receipt is current. Rows P-1 to P-4 and K-1 to K-4 (parallel -composition and the loop grade, 2026-11-19) changed the inputs of the Agda -and Rocq legs: the Agda cone gained three modules and the Rocq file gained -their leg. The Agda and Rocq runs at `15fa113` no longer describe the tree, -so they are `superseded`. The Lean run at `15fa113` was already -`superseded` by resource-algebra#53, which moved the Lean leg and the M-2 -bridge under `lake`. The first `proofs.yml` run on main that contains this -change is recorded next. The 31 rows that were PWG-2i then get the grade -that run supports, and the eight new rows get their first grade. The proofs -of those 31 rows did not change. - -The eight new rows were checked before the pull request in this way, which -is not a receipt: Agda 2.6.3 (`scripts/check-agda.sh`, 16 expected -rejections), Lean 4.35.0 from the pinned release archive -(`scripts/check-lean-leg.sh`, 116 of 116 printed theorems axiom-free) and -Coq 8.18.0 on a copy with `From Stdlib` read as `From Coq` (106 of 106 -printed theorems closed). The pinned Agda 2.6.4.3 and Rocq 9.3.0 run in CI. +All three legs have a current receipt from one `proofs.yml` run on main at +that commit. It is the first run that contains rows P-1 to P-4 and K-1 to +K-4 (parallel composition and the loop grade), and the first run on main of +the Lean leg and the M-2 bridge under `lake`. The eight new rows have their +first grade, and the 31 earlier rows are PWG-2ii again. No row is PWG-1. Since standards#1625 (2026-11-17, owner ruling 2026-11-16), PWG-1 needs every counted witness to name the image its checker @@ -57,7 +44,7 @@ same evidence now gives PWG-2ii. The proofs did not change. M-2 stays ungraded whatever the receipts say. M-2 is `PROVEN` by a Lean bridge only. Its statement names a Lean-only object, so it has no Agda witness and cannot -reach PWG-3 (§8.1). Every other row is graded by the receipts once one is current. No row rests +reach PWG-3 (§8.1). Every other row is graded by the current receipt. No row rests on an axiom: the Agda leg is `--safe` with the trusted base below, the Rocq leg prints every theorem closed under the global context, and the Lean leg prints no axiom for any theorem. @@ -125,10 +112,9 @@ Every theorem row carries this record, except M-2 (see below). E-1, E-2 and E-3 are covered: their Rocq and Lean legs are re-proved by hand, in the same files and by the same routes as every other row, their statements are reconciled (see Statement correspondence), and the receipt below records a -run of all three legs on a commit that contains them. The Lean run of -that receipt is now `superseded` (see Census). -P-1 to P-4 and K-1 to K-4 are re-proved in both legs in the same way. No -receipt records a run of them yet (see Census). +run of all three legs on a commit that contains them. +P-1 to P-4 and K-1 to K-4 are re-proved in both legs in the same way, and +the same receipt covers them. [cols="1,1,1,1,2",options="header"] |=== @@ -138,8 +124,7 @@ receipt records a run of them yet (see Census). | source | `PROVEN` | `AGDA-VERIFIED` -| `superseded`: CI run `44209774341`, job `134484857967`. No current -receipt yet +| CI run `44435465068`, job `135263084676` | Rocq | `re-proved` (transport `unavailable`; see below) @@ -147,28 +132,22 @@ receipt yet under the global context), and `rocqchk` rechecks the compiled file | `RECONCILED`: statements compared and residual differences recorded below -| `superseded`: CI run `44209774341`, job `134484858135`. No current -receipt yet +| CI run `44435465068`, job `135263084711` | Lean | `re-proved`, Lean core only, no Mathlib | `PROVEN`: no axiom (`#print axioms` reports none for every theorem) | `RECONCILED`: statements compared and residual differences recorded below -| `superseded`: CI run `44209774341`, job `134484858201` (`lean` called -directly, before the leg moved under `lake`). No current receipt yet +| CI run `44435465068`, job `135263084426` |=== === Receipts -All three rows below are `superseded`: the Lean row by -resource-algebra#53, and the Agda and Rocq rows by the change that added -P-1 to P-4 and K-1 to K-4 (see Census). - All three legs passed in one `proofs.yml` run, -https://github.com/hyperpolymath/resource-algebra/actions/runs/44209774341[`44209774341`], -on source commit `15fa113c7427e2de009eade9148295d08c046b33` (2026-11-18), the -merge of resource-algebra#45. The same run passed the `Proofs verdict` job +https://github.com/hyperpolymath/resource-algebra/actions/runs/44435465068[`44435465068`], +on source commit `201e05bf3ab7acaa4a7a4a2d24f87c561d4e7f23` (2026-11-19), the +merge of resource-algebra#58. The same run passed the `Proofs verdict` job and the coverage-exemption check. The machine-readable record is `verification/receipts/resource-algebra-proofs_praxis.deed` (§8.5.1). Each job's run summary holds its receipt: the checker and its asserted @@ -180,20 +159,20 @@ version, the invocation, the input digest, the result and the axioms. | Agda | Agda 2.6.4.3 (Debian `agda-bin 2.6.4.3-1+b2`) -| `6d3250447bd3271d…b74ae9d` over the import cone of `src/ResourceAlgebra/All.agda` (13 files, listed in the deed's `(inputs-manifest …)`); `--safe`, the postulate scan of +| `2332490d8a1d57df…3ee08cb9` over the import cone of `src/ResourceAlgebra/All.agda` (16 files, listed in the deed's `(inputs-manifest …)`); `--safe`, the postulate scan of `src/`, and the trusted base: `Level`, `lzero`, `lsuc` and `_⊔_` in `Agda.Primitive`, each bound by its `BUILTIN` pragma | Rocq | Rocq 9.3.0, stdlib 9.2.0, then `rocqchk` -| `b52b1348c6cace05…93c8b1b7` over `ResourceAlgebra.v` (file SHA256 -`4989d461674d9f3d…d760582e`); every printed theorem closed under the +| `14a19b0646438269…36666e8e` over `ResourceAlgebra.v` (file SHA256 +`44231e873c1e82b2…7133b46f`); every printed theorem closed under the global context. Image `rocq/rocq-prover:9.3.0@sha256:97a68f23…` | Lean -| Lean 4.35.0, core only, called directly (`superseded`) -| `015e62121942646e…68d9cb6` over `ResourceAlgebra.lean` (file SHA256 `d1f0ee39d9a8af12…521cc9b`), `lean-toolchain`, `scripts/check-lean-bridge.sh` (file SHA256 `2bb154f8c4c197a7…9f60548`, which holds the tropical-types pin) and the -M-2 bridge, with the import of tropical-types at `f2e465c` (import digest `c176b35aaec8d21b…3095e01`); no printed +| Lean 4.35.0 with Lake 5.0.0, core only, under `lake` +| `f977484c50484262…0a6cff74` over `ResourceAlgebra.lean` (file SHA256 `87476318c1eafe51…a3045f19`), its `lakefile.toml` and `lake-manifest.json`, `lean-toolchain`, `scripts/check-lean-leg.sh`, `scripts/check-lean-bridge.sh` (file SHA256 `6b4dd77974515696…e37de14b`, which holds the tropical-types pin) and the +M-2 bridge with its Lake package files, with the import of tropical-types at `f2e465c` (import digest `c176b35aaec8d21b…3095e01`); no printed theorem depends on an axiom, in the leg or in the bridge |=== @@ -229,6 +208,12 @@ kept by GitHub for a limited time. Earlier receipts, all `superseded`: +* Run `44209774341` on source commit +`15fa113c7427e2de009eade9148295d08c046b33` (2026-11-18), the merge of +resource-algebra#45. It derived PWG-2ii for 31 rows. It predates P-1 to P-4 +and K-1 to K-4, and its Lean job called `lean` directly, before the leg +moved under `lake` (resource-algebra#53). + * Run `44202493421` on source commit `2b9bf3fb0b0435705b0ad92b929edfe80a9be000` (2026-11-17), the merge of resource-algebra#48. A hand-written receipt that derived PWG-2ii for 28 rows @@ -460,7 +445,7 @@ Mathlib | `PROVEN`: no axiom (`#print axioms` reports none for any of the six printed theorems) | `LEAN-CHECKED`, not `RECONCILED`: there is no Agda statement to reconcile with -| CI run `44209774341`, job `134484858201`, with +| CI run `44435465068`, job `135263084426`, with `(import :repository "hyperpolymath/tropical-types" :commit "f2e465c…")` (§8.5.1) |=== @@ -498,25 +483,25 @@ counted. | `ResourceAlgebra`: alternatives form an idempotent commutative monoid and sequencing a monoid. | `PROVEN` (the definition, plus the derived `unitʳ`) -| ungraded +| 2ii | C-2 | The relation `≼` induced by ⊕ is a partial order. `𝟘` is least and `⊕` is its join. | `PROVEN` (`≼-refl`, `≼-trans`, `≼-antisym`, `𝟘-least`, `⊕-upperˡ/ʳ`; the least-upper-bound half is `⊕-least`, from main #21) -| ungraded +| 2ii | C-3 | Every dioid is ordered: distributivity makes `⊗` monotone on both sides. | `PROVEN` (`IsDioid.ordered`) -| ungraded +| 2ii | C-4a | A dioid need not be annihilating. HWM is a dioid and is not annihilating. | `PROVEN` (`HWM-dioid`, `HWM-not-annihilating`); control `tests/reject/HWMAnnihilates.agda` -| ungraded +| 2ii | C-4b | An annihilating ordered algebra need not be a dioid. The diamond lattice @@ -525,7 +510,7 @@ ordered and annihilating, and is not a dioid: a ⊗ (b ⊕ c) = a, while (a ⊗ b) ⊕ (a ⊗ c) = 𝟘. | `PROVEN` (`Diamond`, `Diamond-ordered`, `Diamond-annihilating`, `Diamond-not-dioid`); control `tests/reject/DiamondDistributes.agda` -| ungraded +| 2ii |=== == Instances @@ -538,13 +523,13 @@ ordered and annihilating, and is not a dioid: a ⊗ (b ⊕ c) = a, while | `Loss` is a `ResourceAlgebra`, `IsDioid`, `IsOrdered` and `IsAnnihilating`, with ⊕ = min, ⊗ = +, 𝟘 = inf and 𝟙 = fin 0. | `PROVEN` (Instances/Loss.agda) -| ungraded +| 2ii | M-1 | `MaxPlus` is a `ResourceAlgebra`, `IsDioid`, `IsOrdered` and `IsAnnihilating`, with ⊕ = max, ⊗ = +, 𝟘 = bot and 𝟙 = val 0. | `PROVEN` (Instances/MaxPlus.agda) -| ungraded +| 2ii | M-2 | `MaxPlus` agrees with tropical-types' `instance : CommSemiring Tropical` @@ -563,13 +548,13 @@ names a Lean object. | `HWM` (peak, drift, with drift ≤ peak) is a `ResourceAlgebra`, `IsDioid` and `IsOrdered`, and 𝟘 ≡ 𝟙. | `PROVEN` (Instances/HWM.agda) -| ungraded +| 2ii | H-3 | `HWM` has no annihilating zero. | `PROVEN` (`HWM-not-annihilating`); control `tests/reject/HWMAnnihilates.agda` -| ungraded +| 2ii | N-1 | `MinMax` (the bottleneck algebra) is a `ResourceAlgebra`, `IsDioid`, @@ -577,7 +562,7 @@ and `IsOrdered`, and 𝟘 ≡ 𝟙. ⊕ = min and ⊗ = max (`_⊔g_`, with `inf` absorbing), with 𝟘 = inf and 𝟙 = fin 0. | `PROVEN` (Instances/MinMax.agda) -| ungraded +| 2ii | Q-1 | `Questions n` (a "which" grade over `n` downstream questions) is a @@ -586,19 +571,19 @@ and `IsOrdered`, and 𝟘 ≡ 𝟙. ⊕ = intersection and ⊗ = union, bit by bit, with 𝟘 = every question lost and 𝟙 = none lost. | `PROVEN` (Instances/Questions.agda) -| ungraded +| 2ii | Q-2 | With two questions, `collapse` (loses question 1) and `shuffle` (loses question 0) each lose one question and are different grades: a count of lost questions cannot tell them apart. | `PROVEN` (`same-amount-different-loss`) -| ungraded +| 2ii | Q-3 | In sequence, `collapse` then `shuffle` loses both questions. | `PROVEN` (`both-lost`) -| ungraded +| 2ii | Q-4 | As alternatives, `collapse ⊕ shuffle` is 𝟙 (no question lost), while @@ -606,7 +591,7 @@ neither branch is 𝟙. So ⊕ answers "is question i kept on some branch?" and not "does one branch keep every question?". | `PROVEN` (`no-branch-achieves-meet`, with Q-2); control `tests/reject/QuestionsMeetIsABranch.agda` -| ungraded +| 2ii | E-1 | `Evidence` (how strongly a proof-ledger row is supported) is a @@ -619,7 +604,7 @@ operations are ⊕ = the stronger (join) and ⊗ = the weaker (meet), with the meet of `≼` (`⊓e-lowerˡ`, `⊓e-lowerʳ`, `⊓e-greatest`). | `PROVEN` (Instances/Evidence.agda); control `tests/reject/EvidenceArguedBelowTested.agda` -| ungraded +| 2ii |=== == Evidence ledger (Ledger.agda) @@ -651,7 +636,7 @@ through dependencies. | `PROVEN` (`eff-≼-own`, `eff-≼-dep`, `eff-≼-reach`); controls `tests/reject/EvidenceProvenOverTested.agda`, `tests/reject/EvidenceArguedLinkIgnored.agda` -| ungraded +| 2ii | E-3 | Maximality: `eff` respects the ledger, and every assignment of statuses @@ -659,7 +644,7 @@ that respects it is below `eff` row by row. So `eff` is the strongest status the ledger supports, and a recorded status above it is an overstatement. | `PROVEN` (`eff-respects`, `eff-greatest`) -| ungraded +| 2ii |=== The worked ledger in `tests/unit/EvidenceTest.agda` closes the loop on @@ -681,26 +666,26 @@ it maps ⊗ to at most ⊗: that is what transporting a bound needs. | Both notions have identities and compose, and every strict morphism is lax. | `PROVEN` (`id-hom`, `_∘h_`, `id-lax`, `_∘l_`, `lax`) -| ungraded +| 2ii | F-2 | `peak : HWM → MaxPlus` is lax: the peak of a sequence never exceeds the sum of the peaks. | `PROVEN` (`peak-lax`) -| ungraded +| 2ii | F-3 | `peak` is not strict: it does not preserve ⊗, whatever it does to the units. | `PROVEN` (`peak-not-strict`); control `tests/reject/PeakIsStrict.agda` -| ungraded +| 2ii | F-4 | No strict morphism leaves HWM for an algebra with 𝟘 ≢ 𝟙. In particular, none goes to MaxPlus or to Loss. | `PROVEN` (`no-strict-from-HWM`, `no-strict-HWM→MaxPlus`, `no-strict-HWM→Loss`) -| ungraded +| 2ii | F-5 | Exactly two strict morphisms Loss → MaxPlus exist, `zero-test` and @@ -710,7 +695,7 @@ morphism is one of them (`Loss→MaxPlus`). They are distinct (`maps-disagree-at-positive`), and every morphism collapses `fin 1` and `fin 2` (`Loss→MaxPlus-collapses`, `no-injective-Loss→MaxPlus`). -| ungraded +| 2ii | F-6 | Exactly one strict morphism MaxPlus → Loss exists, `support′`. @@ -718,13 +703,13 @@ resource-algebra#3 sketched this direction as "symmetric" to Loss → MaxPlus, but it is not: there is one map, not two. The issue's other claim, that every positive grade goes to a single point, holds. | `PROVEN` (`support′-hom`, `MaxPlus→Loss`) -| ungraded +| 2ii | F-7 | The identity on the carrier is a lax morphism Loss → MinMax: a bottleneck never exceeds the sum. | `PROVEN` (`loss-bottleneck`) -| ungraded +| 2ii | F-8 | Out of one question there is exactly one strict morphism into an algebra @@ -732,20 +717,20 @@ never exceeds the sum. 𝟘 ⊗ 𝟘 ≡ 𝟘, which every annihilating algebra has, and HWM too. | `PROVEN` (`one-question-hom`, `one-question-needs`, `one-question-unique`) -| ungraded +| 2ii | F-9 | Taking the first question of `Questions (suc n)`, and dropping it, are strict morphisms. Composed, they read any single question alone. | `PROVEN` (`first-question`, `later-questions`) -| ungraded +| 2ii | F-10 | The number of lost questions does not preserve ⊕: `collapse` and `shuffle` each count 1 and their ⊕ counts 0. So no map into Loss or MinMax that factors through the count is strict or lax. | `PROVEN` (`count-not-⊕`) -| ungraded +| 2ii |=== == Separation (Separation.agda) @@ -763,26 +748,26 @@ for checks over a bounded or sampled domain. | Sequencing peaks 2 and 3 gives 3 under HWM and 5 under MaxPlus: an occupancy bound is not a cost bound. | `PROVEN` (`occupancy-is-not-cost`) -| ungraded +| 2ii | S-2 | On the same numbers, alternatives give `fin 2` under Loss and `val 3` under MaxPlus: echo's join is min, while tropical's is max. | `PROVEN` (`alternatives-differ`) -| ungraded +| 2ii | S-3 | `val 1 ≼ val 2` in MaxPlus but `fin 2 ≼ fin 1` in Loss: the induced orders run in opposite directions. | `PROVEN` (`maxplus-order`, `loss-order-reversed`) -| ungraded +| 2ii | S-4 | Sequencing 2 then 3 gives `fin 5` under Loss but `fin 3` under MinMax: a bottleneck bound is not a loss bound. | `PROVEN` (`bottleneck-is-not-sum`); `tests/reject/MinMaxIsLoss.agda` refuses the general identification -| ungraded +| 2ii |=== == Parallel composition (Parallel.agda) @@ -807,14 +792,14 @@ right and is monotone on both sides. A dioid with a commutative ⊗ is parallel with `_∥_` = ⊗, and then both exchange laws hold, as an equality. | `PROVEN` (`IsParallel` with `unitʳ`, `distribʳ`, `monoˡ`, `monoʳ`; `seq-parallel`, `seq-sync-below`, `seq-sync-above`) -| ungraded +| 2ii | P-2 | Work. `Loss` and `MaxPlus` are parallel with `_∥_` = + (their ⊗), and have both exchange laws. | `PROVEN` (`Loss-parallel`, `Loss-sync-below`, `Loss-sync-above`, `MaxPlus-work`, `MaxPlus-work-sync-below`, `MaxPlus-work-sync-above`) -| ungraded +| 2ii | P-3 | Span. `MaxPlus` is parallel with `_∥_` = max on finite costs (`bot` @@ -823,7 +808,7 @@ absorbs). `SyncAbove` holds: max (a + c) (b + d) ≤ max a b + max c d. middle and 1 when they run free. | `PROVEN` (`MaxPlus-span`, `MaxPlus-span-sync-above`, `MaxPlus-span-not-sync-below`); control `tests/reject/SpanSyncBelow.agda` -| ungraded +| 2ii | P-4 | Occupancy. `HWM` is parallel with `_∥_` = the sum of the peaks and the @@ -832,7 +817,7 @@ peak 1, one before the meeting and one after it, hold 1 when they meet in the middle and are graded 2 when they run free. | `PROVEN` (`HWM-parallel`, `HWM-sync-below`, `HWM-not-sync-above`); control `tests/reject/HWMSyncAbove.agda` -| ungraded +| 2ii |=== P-3 and P-4 together say that elapsed time and occupancy exchange in @@ -854,7 +839,7 @@ greatest grade, `star a` = `𝟙` for every `a` is a star. `Loss` is such an algebra, so a loop grades as `fin 0`. | `PROVEN` (`IsStar` with `star-𝟙`, `star-step`; `unit-top-star`, `Loss-unit-top`, `Loss-star`) -| ungraded +| 2ii | K-2 | A grade whose loop equation 𝟙 ⊕ (a ⊗ s) ≡ s has no solution rules a @@ -864,7 +849,7 @@ carrier. | `PROVEN` (`no-star`, `MaxPlus-no-fixpoint`, `MaxPlus-no-star`, `HWM-no-fixpoint`, `HWM-no-star`); control `tests/reject/MaxPlusLoopBounded.agda` -| ungraded +| 2ii | K-3 | `MaxPlusTop` (`Instances/MaxPlusTop.agda`: ℕ with −∞ and +∞, ⊕ = max, @@ -873,13 +858,13 @@ carrier. embeds in it by a strict, injective morphism. | `PROVEN` (`MaxPlusTop`, `MaxPlusTop-dioid`, `MaxPlusTop-ordered`, `MaxPlusTop-annihilating`, `lift-hom`, `lift-injective`) -| ungraded +| 2ii | K-4 | `MaxPlusTop` has a star: `bot` and `val 0` loop to `val 0`, and every dearer body, and `top`, loop to `top`. | `PROVEN` (`star-t`, `MaxPlusTop-star`) -| ungraded +| 2ii |=== == Correspondences with other repositories (not graded) diff --git a/docs/status/ROADMAP.adoc b/docs/status/ROADMAP.adoc index 195d37f..f3c75ac 100644 --- a/docs/status/ROADMAP.adoc +++ b/docs/status/ROADMAP.adoc @@ -28,7 +28,7 @@ Initial development phase. * [ ] Pin the digests in `.github/workflows/proofs.yml`, print each image digest in the job summaries, and record a new receipt from a run on the merge commit. -* Until then the 28 rows stay PWG-2ii. Detail and rejected alternatives: +* Until then the 39 graded rows stay PWG-2ii. Detail and rejected alternatives: `docs/status/PROOF-NEEDS.adoc`, PN-7. == Future Directions diff --git a/scripts/proof-receipt.sh b/scripts/proof-receipt.sh index 70fd033..7c45b88 100755 --- a/scripts/proof-receipt.sh +++ b/scripts/proof-receipt.sh @@ -103,32 +103,15 @@ receipt_claims() { # images pinned by digest and install the checker at run time, which the # owner ruled on 2026-11-18 is not isolation (docs/status/PROOF-NEEDS.adoc # PN-7): their rows name no image, and no row can reach PWG-1 until they -# run in prover images. The Lean row is superseded: the leg and the bridge -# moved under lake, and the first run on main under lake replaces it. -# Earlier runs are listed, superseded, in docs/status/PROOF-STATUS.adoc. +# run in prover images. The three rows are the jobs of proofs.yml run +# 44435465068 on main at 201e05b, the merge of resource-algebra#58; each was +# printed by --run-row. Earlier runs are listed, superseded, in +# docs/status/PROOF-STATUS.adoc. receipt_runs() { - local u=https://github.com/hyperpolymath/resource-algebra/actions/runs/44209774341/job - local c=15fa113c7427e2de009eade9148295d08c046b33 - # The Lean invocation of that run, before the legs moved under lake. - local lean_direct='lean ResourceAlgebra.lean in verification/proofs/lean4, core only; then bash scripts/check-lean-bridge.sh (lean -o on TropicalSessionTypes.lean and ResourceAlgebra.lean, then lean TropicalBridge.lean with LEAN_PATH)' - cat <