Repository navigation
Unsealed specs, do not compile: undeclared identifiers: batch 3 of 9 (21 specs) #8396
Description
Part of #8393 (epic #8095). These specs do not compile: the generated Zig uses an identifier nothing declares. Without a seal, t27c frontier can never reuse their verdict.
| spec | first compile error (t27c test-report, 2026-10-10) |
|---|---|
specs/igla/training/scale_up.t27 |
error: use of undeclared identifier 'latency_ms' |
specs/interop/gf_cross_language.t27 |
error: use of undeclared identifier 'gf32_from_f64' |
specs/isa/ternary_pattern_matching.t27 |
error: use of undeclared identifier 'eq' |
specs/isa/ternary_search.t27 |
error: use of undeclared identifier 'eq' |
specs/isa/ternary_sorting.t27 |
error: use of undeclared identifier 'eq' |
specs/isa/ternary_tree.t27 |
error: use of undeclared identifier 'not' |
specs/math/e8_lie_algebra.t27 |
error: use of undeclared identifier 'PHI_lv' |
specs/math/radix_economy.t27 |
error: use of undeclared identifier 'results' |
specs/math/sacred_physics.t27 |
error: use of undeclared identifier 'PI_lv' |
specs/memory/formula_embed.t27 |
error: use of undeclared identifier 'Float' |
specs/memory/memory_primitives.t27 |
error: use of undeclared identifier 'not' |
specs/memory/semantic_search.t27 |
error: use of undeclared identifier 'Float' |
specs/ml/activation/elu_activation.t27 |
error: use of undeclared identifier 'math' |
specs/ml/activation/gelu_approx_activation.t27 |
error: use of undeclared identifier 'math' |
specs/ml/activation/leaky_relu_activation.t27 |
error: use of undeclared identifier 'alloc' |
specs/ml/activation/silu_swish_activation.t27 |
error: use of undeclared identifier 'exp' |
specs/ml/activation/softmax.t27 |
error: use of undeclared identifier 'range' |
specs/ml/activation/tanh_activation.t27 |
error: use of undeclared identifier 'math' |
specs/ml/igla_champion_capsule.t27 |
error: use of undeclared identifier 'elapsed_time_ns' |
specs/ml/layers/batchnorm_layer.t27 |
error: use of undeclared identifier 'valid_input' |
specs/ml/layers/dense_layer.t27 |
error: use of undeclared identifier 'valid_input' |
Attribute an error to the file it is in: an identifier that comes from an import is fixed in the import or the imported spec.
Acceptance criteria
t27c test-report <spec>prints nodoes not compileline for any spec under Boundary.- With the line above satisfied, every spec under Boundary reports
FAIL 0and at least one pass, and no test or invariant is deleted to get there. - Every spec under Boundary has a v2 seal written by
t27c seal <spec> --save:seal_schema 2,tests.failed 0, committed under.trinity/seals/. - A defect in the codegen is filed on epic: self-host t27c -- rewrite the compiler from Rust into t27 #5980 with a minimal
.t27repro and linked here; the spec is not rewritten to dodge it.
Boundary
specs/igla/training/scale_up.t27
specs/interop/gf_cross_language.t27
specs/isa/ternary_pattern_matching.t27
specs/isa/ternary_search.t27
specs/isa/ternary_sorting.t27
specs/isa/ternary_tree.t27
specs/math/e8_lie_algebra.t27
specs/math/radix_economy.t27
specs/math/sacred_physics.t27
specs/memory/formula_embed.t27
specs/memory/memory_primitives.t27
specs/memory/semantic_search.t27
specs/ml/activation/elu_activation.t27
specs/ml/activation/gelu_approx_activation.t27
specs/ml/activation/leaky_relu_activation.t27
specs/ml/activation/silu_swish_activation.t27
specs/ml/activation/softmax.t27
specs/ml/activation/tanh_activation.t27
specs/ml/igla_champion_capsule.t27
specs/ml/layers/batchnorm_layer.t27
specs/ml/layers/dense_layer.t27
.trinity/seals/
User Scenarios
- Given the 21 specs listed in Boundary each fail to compile with an undeclared identifier error, When the missing declarations are added or the codegen defect is filed with a minimal repro, Then
t27c test-report <spec>prints no "does not compile" line for any of those specs. - Given all 21 specs compile cleanly, When
t27c test-report <spec>is run for each, Then every spec reportsFAIL 0with at least one passing test and no test or invariant was deleted to achieve it. - Given the specs compile and pass, When
t27c seal <spec> --saveis run for each, Then a v2 seal is written under.trinity/seals/withseal_schema 2andtests.failed 0.
Requirements
FR-001: Each of the 21 specs under Boundary MUST compile without undeclared identifier errors when running t27c test-report <spec>.
FR-002: For each spec, the fix MUST be a declaration in the spec or its imports, not a deletion of any test or invariant.
FR-003: If an undeclared identifier originates from a codegen defect, a minimal .t27 repro MUST be filed on issue #5980 and linked from this issue; the spec MUST NOT be rewritten to dodge the defect.
FR-004: After FR-001 is satisfied, every spec under Boundary MUST report FAIL 0 and at least one pass under t27c test-report <spec>.
FR-005: For each spec under Boundary, t27c seal <spec> --save MUST produce a v2 seal file under .trinity/seals/ with seal_schema 2 and tests.failed 0.
Success Criteria
t27c test-report specs/igla/training/scale_up.t27prints no "does not compile" line and exits 0t27c test-report specs/interop/gf_cross_language.t27prints no "does not compile" line and exits 0t27c test-report specs/isa/ternary_pattern_matching.t27prints no "does not compile" line and exits 0t27c test-report specs/isa/ternary_search.t27prints no "does not compile" line and exits 0t27c test-report specs/isa/ternary_sorting.t27prints no "does not compile" line and exits 0t27c test-report specs/isa/ternary_tree.t27prints no "does not compile" line and exits 0t27c test-report specs/math/e8_lie_algebra.t27prints no "does not compile" line and exits 0t27c test-report specs/math/radix_economy.t27prints no "does not compile" line and exits 0t27c test-report specs/math/sacred_physics.t27prints no "does not compile" line and exits 0t27c test-report specs/memory/formula_embed.t27prints no "does not compile" line and exits 0t27c test-report specs/memory/memory_primitives.t27prints no "does not compile" line and exits 0t27c test-report specs/memory/semantic_search.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/elu_activation.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/gelu_approx_activation.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/leaky_relu_activation.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/silu_swish_activation.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/softmax.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/activation/tanh_activation.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/igla_champion_capsule.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/layers/batchnorm_layer.t27prints no "does not compile" line and exits 0t27c test-report specs/ml/layers/dense_layer.t27prints no "does not compile" line and exits 0- Each of the 21 specs reports
FAIL 0and at least 1 pass undert27c test-report - 21 v2 seal files exist under
.trinity/seals/withseal_schema 2andtests.failed 0
Activity
gHashTag commentedon Oct 9, 2026
Priority:
specs/math/sacred_physics.t27is the largest missing hub int27c frontier. 26 specs are held in closure rebuild behind it, so fix it first.PI_lv, is a compiler defect: gen-zig: a shadowing local's _lv rename reaches its own initializer (const PI_lv = PI_lv) #8416, with the fix in gen-zig: an empty typed array &[_]T{} lowers to .{} (Closes #8390) #8415.sin2_theta12_juno_2025_compatibilitycallssin2_theta12_trinity(), and nothing declares it.8*phi^4/3/pi = 0.30906..., but that expression evaluates to 5.81794, not 0.309. Do not implement the function from that comment. The claim (C-phi-005) needs its real formula from the author. Until then, the honest fix is to remove the test or mark it unverified, with the reason written down. Do not invent a formula that happens to hit 0.309.gHashTag commentedon Oct 10, 2026
Correction to my priority note above. "26 specs held behind it" counted every dependent in closure rebuild, including dependents with blockers of their own. Counting only dependents whose seal passes and whose ONLY non-reusable import is
sacred_physics.t27, fixing it frees 3 specs, not 26. That is still the top hub by that measure; the next ones free 1 each. Thesin2_theta12_trinityquestion still needs the author.