The existing inverse_intent_resolves_one_admitted_transition_after_restart fixture fails while reopening the WAL after an inverse replacement has applied. The new history host refuses recovery with Wal(Recovery(Validation(LsnContinuityMismatch))) at crates/warp-core/tests/external_consumer_contract_fixture_tests.rs:687.
This is independently reproducible before Bunny integration: unchanged base 5f99097d9a5c45a91ab22ec996f7a24266f92a13 fails using its original Rust1.90.0, and also with1.96.0, in the same guarded Docker worker. The original-toolchain run completed compilation, ran exactly the named test, and reported0passed/1failed, Cargo101. The test itself is unchanged by companionPR750; its only WAL-file production changes are equivalent strict-lint corrections. The root cause is not yet established.
Reproduction inside the bounded Docker validation environment, on the pinned base:
cargo +1.90.0 test --locked -p warp-core \
--features native_rule_bootstrap,trusted_runtime,host_test \
--test external_consumer_contract_fixture_tests \
inverse_intent_resolves_one_admitted_transition_after_restart -- --exact
Observable outcome: a committed inverse transition remains recoverable after the next restart, with valid contiguous WAL evidence and the expected typed inverse derivation.
Scope: isolate and repair the WAL append/recovery or fixture defect that produces this failure. Preserve transaction integrity, causal parent binding and refusal of genuinely corrupt/noncontiguous history. Do not suppress continuity validation or merely remove the restart assertion. Excludes Bunny arithmetic, fixed-point language features, application-specific Jim operations and unrelated WAL redesign.
Acceptance checks:
- Preserve the existing deterministic failing fixture as RED on the pinned base, then make it pass through both restart boundaries.
- Verify recovered document value, frontier/provenance advancement and typed inverse receipt derivation.
- Retain meaningful negative coverage that refuses actual LSN gaps/duplicates or malformed transaction evidence.
- Run the relevant WAL/runtime suites and selected external-consumer feature lane in guarded Docker, with compiler/source identities recorded.
Related tracking scope: inverse Intent path #494 describes the broader capability. This issue isolates a concrete restart failure in the existing implementation and does not duplicate that full feature scope.
No prerequisite is currently established; investigation may identify one with evidence. This should be one independently mergeable repair that leaves ordinary admission/recovery checks working. The Bunny foundation is not a prerequisite; this issue records a pre-existing failure observed during PR750 validation.
Evidence: retained echo-bunny-echo-final-gates.log, baseline-original-toolchain section at base5f99097d, and earlier echo-bunny-echo-verify2.log original-base1.96 comparison. The complete final log hash will be attached after the ongoing candidate gate finishes. These are process-level restart witnesses; no power-loss claim is made.
The existing
inverse_intent_resolves_one_admitted_transition_after_restartfixture fails while reopening the WAL after an inverse replacement has applied. The new history host refuses recovery withWal(Recovery(Validation(LsnContinuityMismatch)))atcrates/warp-core/tests/external_consumer_contract_fixture_tests.rs:687.This is independently reproducible before Bunny integration: unchanged base
5f99097d9a5c45a91ab22ec996f7a24266f92a13fails using its original Rust1.90.0, and also with1.96.0, in the same guarded Docker worker. The original-toolchain run completed compilation, ran exactly the named test, and reported0passed/1failed, Cargo101. The test itself is unchanged by companionPR750; its only WAL-file production changes are equivalent strict-lint corrections. The root cause is not yet established.Reproduction inside the bounded Docker validation environment, on the pinned base:
cargo +1.90.0 test --locked -p warp-core \ --features native_rule_bootstrap,trusted_runtime,host_test \ --test external_consumer_contract_fixture_tests \ inverse_intent_resolves_one_admitted_transition_after_restart -- --exactObservable outcome: a committed inverse transition remains recoverable after the next restart, with valid contiguous WAL evidence and the expected typed inverse derivation.
Scope: isolate and repair the WAL append/recovery or fixture defect that produces this failure. Preserve transaction integrity, causal parent binding and refusal of genuinely corrupt/noncontiguous history. Do not suppress continuity validation or merely remove the restart assertion. Excludes Bunny arithmetic, fixed-point language features, application-specific Jim operations and unrelated WAL redesign.
Acceptance checks:
Related tracking scope: inverse Intent path #494 describes the broader capability. This issue isolates a concrete restart failure in the existing implementation and does not duplicate that full feature scope.
No prerequisite is currently established; investigation may identify one with evidence. This should be one independently mergeable repair that leaves ordinary admission/recovery checks working. The Bunny foundation is not a prerequisite; this issue records a pre-existing failure observed during PR750 validation.
Evidence: retained
echo-bunny-echo-final-gates.log, baseline-original-toolchain section at base5f99097d, and earlierecho-bunny-echo-verify2.logoriginal-base1.96 comparison. The complete final log hash will be attached after the ongoing candidate gate finishes. These are process-level restart witnesses; no power-loss claim is made.