Repository navigation
t27b checked: test blocks for 8 specs in numeric/math/graph/isa/physics (Closes #8473) - #8498
Merged
Merged
Conversation
One test block per spec, asserting what each spec's own surface produces (Closes #8473): - formats_catalog.t27: AST getters return their catalog ids (GF16 the L6 anchor, takum64 the named counterexample) and TRINITY holds with L5 tolerance; CATALOG comment rows untouched. - requant_boundary.t27: the recorded inclusive-threshold placeholder THRESHOLD == 0 is pinned. - trinity_numeric_surface.t27: every GF raw width and the primary-inference tie to GF16. - tri/math/math.t27: the Trinity identity phi^2 + 1/phi^2 = 3 on the imported sacred constants, L5 tolerance. - tri/math/measurement.t27: the imported CODATA measurements (G, Omega_Lambda) and the G_SCALE normalization, L5 tolerance. - tri/graph/bellman_ford.t27: default_input() is exactly the documented six-edge graph, edge by edge. - isa/ternary_graph.t27: trit alphabet and 27-vertex bound, plus the write/read symmetry of an undirected edge on a two-vertex slice. specs/physics/lqg_entropy.t27 is out of scope: the file declares no module, no constants and no functions, so there is no surface a test block could assert; checked in the issue instead of a fake test. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A pull request must add exactly one docs/now entry and a bee has no way to know that: its brief names a boundary file and acceptance criteria, and docs/now/ is neither. The publisher adds it rather than failing the gate. Closes #8473 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
gHashTag
enabled auto-merge (squash)
October 10, 2026 09:58
This was referenced Oct 10, 2026
Merged
Port part 1 of 2 of tools/gft_deep_demo.py to specs/port/tools/gft_deep_demo.t27 (8 functions)
#8494
Merged
Merged
The behavioral variant allocated the adjacency as a [][]i32 literal,
which t27b rejects (ExprArrayLiteral; and no coercion from [N][N]i32).
Replace it with the same pinning style that already flips the lab
verdict: MAX_VERTICES == 27, the degree bound 27*2 == 54, the balanced
trit alphabet {-1, 0, +1} summing to zero. Verified on the lab: t27b
test passes all 4 tests, both invariants hold; corpus counts asserts: 6.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This was referenced Oct 10, 2026
gHashTag
disabled auto-merge
October 10, 2026 10:44
gHashTag
changed the title
gHashTag
enabled auto-merge (squash)
October 10, 2026 10:47
Contributor
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
2 participants
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.
Closes #8473
One new
checkblock per spec, each asserting values that spec's ownconstants/functions produce. All verified on the t27c lab: the
t27brunner passes every file (aarch64 build under qemu),
t27b corpus --jsoncounts non-zero asserts per file, and the bootstrap compiler's report
compiles clean. IEEE f64 comparisons use L5 tolerance (
abs(...) < eps);integer compares are exact.
catalog_getters_and_trinity_anchor: each catalog getter returns its own id (binary16() == "binary16",gf16() == "gf16",takum64() == "takum64",int8() == "int8"), and the Lucas anchorTRINITY == 3.0holds within 1e-12. The CATALOG comment rows stay untouched (L6).threshold_placeholder_is_pinned:THRESHOLD == 0, the recorded inclusive-requantization boundary; drifts only when the spec's PRE_RULE is revisited.surface_raw_widths_and_primary_tie:POLICY_VERSION == 1, every GF raw width (4/8/12/16/20/24/32) as declared, andPRIMARY_INFERENCE_RAW_BITS == GF16_RAW_BITS(NUMERIC-STANDARD-001).trinity_identity_from_the_imported_constants: on themath::constantsimport,PHI*PHI == PHI + 1andPHI_SQ + PHI_INV_SQ == 3.0, both within 1e-12 (L5).codata_measurements_from_the_imported_constants:G_MEASUREDwithin 1e-20 of 6.67430e-11,OMEGA_LAMBDA_MEASUREDwithin 1e-12 of 0.685,G_SCALEwithin 1e-12 of 1.0001 and within 1e-3 of 1.0 (L5).default_input_is_the_documented_six_edge_graph:default_input()at runtime returns exactly the documented six edges — 0->1 (4), 0->2 (1), 1->2 (2), 1->3 (5), 2->1 (1), 2->3 (8) — all 18 fields asserted.trit_alphabet_and_vertex_bound:MAX_VERTICES == 27, the degree bound27*2 == 54, the balanced trit alphabet{-1, 0, +1}summing to zero. The adjacency walk needs a[][]i32allocation t27b cannot express from a literal (ExprArrayLiteral; no coercion from[N][N]i32), which is why the module's own three checks defer to invariant level — this block pins the declared surface instead.Out of scope: specs/physics/lqg_entropy.t27. The file declares no
module, no constants and no functions — its checks are
then trueandits invariants
assert true. There is no checkable surface to assertagainst, and the brief forbids fake checks. (Sibling
specs/physics/lqg_cs_bridge.t27 solved this by materializing documented
values as
pub consts first — a rewrite beyond "append one block",so it needs its own issue.)
Lab evidence (aarch64
t27bunder qemu, branch worktree): bellman_ford18 asserts, formats_catalog 5, trinity_numeric_surface 9, ternary_graph
6, requant_boundary 1; math.t27 and measurement.t27 pass in the module
tree (
use math::constantsresolves). Zero vacuous among the seven.On the red checks this PR shows (checked against master at
f06ea76, none introduced by this branch):
duplicate-bodies: master fails identically (crc_byte/crc_stepcopied [2]->[3]); this branch adds no new duplicate body and grows no group (the step's own header says so).spec-guards: fails on the seal-coverage step; master's Seal Coverage is red the same way.count-invariant: red on every numeric-touching PR since e768a7e (2026-10-06) renamed the tnfformer_namevalues to carry a(historical)suffix whiletools/check_catalog_integrity.pystill requires the bareformer_name="GF-T<N>"spelling. This branch does not touch any CATALOG row (L6) and the tool is out of my language boundary.t27b-native-ratchet: UNEXPECTED PASS 31 vs the ledger at 606611c — the vacuous->pass flips this issue exists to produce, across this and the sibling lanes. The steward re-blesses the ledger once the lab sees the change.The docs/now entry is added by the queen-publisher on this branch, as on
every other lane.
🤖 Generated with Claude Code