Repository navigation
The type-parameter-in-value-position wall fires: flip its pinned hole to a firing red (split of #11819, parts 1+3) - #12163
Merged
Conversation
… to a firing red
gunbc#12045's regen carried v1.compiler.infer's declared-roster predicate into the seed, so
TypeParameterInValuePosition now refuses `fn any_param<T>(x: T) -> Int { T }` at its own
location. The claim that pinned the hole at == 0 was therefore silently red on main
(claim_batch on a clean ada85b9: FAIL, with the generic-fn and named-non-inhabitant controls
PASS on the same run).
- test.claim.type_argument_kind_inhabitance_witness_test: the pinned hole becomes
a_type_parameter_in_value_position_must_refuse (>= 1; the -1 not-runnable arm stays red),
paired with the existing generic-fn positive control.
- gunbc.kind_reflection_seed_growth retired by its own trigger: the stub it cited
(binding_resolves_to_type_parameter) no longer exists in the seed, and the refusal fires.
- gunbc.recurring_failure_mode a_wall_declared_in_dag_is_inert_in_the_seed_that_compiles:
evidence repointed, plus a receipt that the climb was as silent as the gap.
Split part (1)+(3) of superseded gunbc#11819.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…instead of retiring it (review 70608) Its trigger also named the four-parameter kind-inhabitance fold. The seed still carries the three-parameter type_arg_kind_inhabitance (no kind_inhabitant_matches_resolved) and the single-arm WidthResolution alias, so only the value-position half was discharged by #12045. The row now cites those two seed declarations, with the trigger restated for them. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Re review 70608: confirmed and fixed in 5fd0d04. The seed still carries the three-parameter — sent from swift-bat-511 |
… pinned defect (review 70618) Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Re review 70618: confirmed and fixed. The stale header line above — sent from swift-bat-511 |
briansrls
approved these changes
Sep 24, 2026
briansrls
left a comment
briansrls
left a comment
Contributor
There was a problem hiding this comment.
APPROVE-SOURCE at exact head 3e928815dc8b160b20cb5b1da93eb01a801e1640.
The split is correctly bounded:
- The formerly pinned silence is flipped to the firing
TypeParameterInValuePositionRED, with the generic-function control retained and the not-runnable-1disposition unable to satisfy the claim. kind_reflection_seed_growthis narrowed rather than prematurely retired. It stops citing the discharged literal-false stub and continues to carry the two actual remaining divergences: the seed's three-parametertype_arg_kind_inhabitancewithoutkind_inhabitant_matches_resolved, and the single-armWidthResolutionalias versus the declared coproduct.- The trigger is correspondingly narrowed to regeneration of those exact authorities.
- The recurring-failure receipt records that the wall's climb was silent and makes running pinned holes part of the lesson for future seed regenerations.
MachineWidth<N>reification and native-emission evidence remain explicitly out of scope for P2.
All five required jobs are green at this exact head. No further source repair is requested. A head move or main rebind requires a delta review.
gunbai-bot Bot
pushed a commit
that referenced
this pull request
Oct 2, 2026
…arrier or dropped with a located reason; kind_reflection row discharged The stale-justification census went 97 -> 3 at identity grain against the current tree. Three classes of edit: REPAIRED (~44 DeclarationRefs): citations that named an enclosing module while the subject had moved, re-pathed to the carrier that now produces it - target_invocation: 15 compile_clean refs v1_compiler.cli_run -> v1_compiler.cli_run.compile_clean (the subjects live in the inline mod), plus the render ref dropped (impl method on QualificationRefusal, floor_memory_supervisor.rs:93 -- uncitable, not dead) - namespace bridges: v1_interpreter -> v1_compiler.v1_interpreter (the module moved into the generated infer/resolve carrier) - generated_artifact_boundary: 5 refs -> v1_compiler.generated_artifact_ boundary_host.tests (subjects in the inline mod tests at :443) - bare_reference_scanner module_self_declared_names -> v1_compiler.cli_run.bare_reference_scanner_tests - floor_cost_debt: citation repaired to run_required_floor (the constant became fn-local inside it at required_floor_runner.rs:11968) -- repaired rather than deleted because that row's trigger has NOT fired and repair-to-live-carrier is the teardown precedent - floor_teardown_attribution: citation re-inserted as v1_compiler.cli_run::release_process_caches_at_exit -- the old subject was RENAMED, not deleted (b1f4f11), and the live fn is called from claim_executor.rs:2435; current_boundary sentence corrected to what the one whole-run teardown figure still measures DROPPED (~25 citations, each with a 2-line doctrinal comment): subjects that are macro-declared extents (thread_local! stores in floor_memory_instrumentation, modeled_operation_realization, cross_claim_pure_share, cross_claim_demand_census -- gunbc.rust_item_scan MacroScope cannot name them at item grain, so the citation can never be anything but stale noise) or impl methods with no item-grain identity. DELETED: kind_reflection_seed_growth.dag, file + roster entry + import. Its regen trigger FIRED: both cited subjects now carry the four-param fn with kind_inhabitant_matches_resolved x2 and the two-arm enum, regenerated by 3150d9b #12805 -- the row was 'narrowed, not retired' per #12163 until that regen carried the second half, and deletion IS its discharge. KEPT (3, per operator ruling): v1.compiler.emit_rust:: emit_source_root_eval_driver_main_rs, v1_compiler.main::RetainedCliHost, v1_compiler.v1_interpreter::write_file_create_new -- each subject moved to a generated mirror, neither row's trigger has fired, and the census's stale join compares against the live hand-Rust population only, so it cannot address a generated subject. Typed dispositions in the PR body. Admission COMPLETE; well-formedness gate green; the 87-row roster folds with no empty declaration list and no duplicate display key (1210 declarations); duplicate-justifications 0; uncitable-items 1 (the declared Display impl, uncitable by rule). Witnesses: dag/test/claim/seed_growth_admission_witness_test.dag answers 7/7 PASS, and the census instrument exits 0 with stale-justifications (3).
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.
Split parts (1) and (3) of the superseded #11819 (closed; its claims are mapped in the closing comment). Part of the App Attest native-crypto node.
What was true on main
#12045 regenerated
v1_compiler_infer.rsfrom the.dag. That removed thebinding_resolves_to_type_parameter = falsestub and carriedname_is_enclosing_declared_type_parameterandInferScope.enclosing_declared_type_param_namesinto the seed. SoTypeParameterInValuePositionnow fires, and the claim that pinned it as an inert hole (== 0) was silently red on main:git clean -fdx)ada85b9, 0 changed files:a_type_parameter_in_value_position_does_not_yet_refuse_and_this_pins_that_holea_generic_function_not_misusing_its_parameter_still_compilesa_named_non_inhabitant_at_a_kinded_position_must_refuse(proof the harness reaches the check)Change
test.claim.type_argument_kind_inhabitance_witness_test: the pinned hole becomesa_type_parameter_in_value_position_must_refuse(>= 1). The-1couldn't-run arm stays red, so a harness that never reaches the check can't pass it. Positive control: the existing generic-fn claim. It stays as the permanent regression control (DESIGN §4b(4)).gunbc.kind_reflection_seed_growth. Its trigger had two halves. The value-position half is discharged (the stub is gone and the refusal fires). The four-parameter kind-inhabitance fold is NOT: the seed still has the three-parametertype_arg_kind_inhabitancewith nokind_inhabitant_matches_resolved, andpub type WidthResolution = PointerWidth. The row now cites those two seed declarations, with the trigger restated for them, and stays onseed_growth_justification_roster.gunbc.recurring_failure_modea_wall_declared_in_dag_is_inert_in_the_seed_that_compiles: evidence repointed to the new claim, plus a receipt that the climb was as silent as the gap, and that the first draft of this PR repeated the class one level up. A lane that regenerates the seed should run the pinned holes over the modules it regenerated.Evidence at this head (
5fd0d04, clean tree, 0 changed files)type_argument_kind_inhabitance_witness_test: all 6 claims PASS. That includes the new red, the generic-fn control, both wrong-kind negatives and both admitted-inhabitant positives.seed_growth_admission_witness_test: all 7 claims PASS, including the roster-membership and stale-row checks over the narrowed row. Scope: that is the witness file only; the stale check over the real seed surface runs where CI runs the seed-growth gate.Not in scope (P2): actual
MachineWidth<N>reification plus a native-emission receipt.std.machine_constraintsmachine_width_literal_roster_frontierstays a stated frontier.🤖 Generated with Claude Code