feat(epic4): безопасные повторы и отмена durable Task (Story 4.4)
CI / UI Checks (pull_request) Successful in 8s
CI / Rust Checks (pull_request) Successful in 20m46s
CI / Rust Feature Validation (pull_request) Skipped
CI / Frontend E2E (pull_request) Skipped
CI / Community Image Smoke (pull_request) Skipped
CI / Required Checks (pull_request) Successful in 1s
CI / Deploy (pull_request) Skipped

Durable per-outbound attempt до отправки запроса, bounded retry/read,
server-owned verified writes, честный cancel/unknown, deadline/fencing,
additive schema v29. Staged renames task.rs/tasks.rs разложены на модули.

Includes disclosed cosmetic parity exception: единственная правка вне
принятой дельты — удалена лишняя пустая строка в конце
scripts/ci-story-checks.sh (EOF blank line); story-код не менялся.
This commit is contained in:
bsodfather committed 2026-10-30 12:21:57 +00:00
1 parent 8c02380762
commit a3057de51e
44 files changed
+4547 -815

No files matched your search

+7
View File
@@ -85,3 +85,10 @@ run_atdd "durable task migration" -p crank-registry --test integration migration
# claims before a slow HTTP drain, not after it) and one child attempt span per adapter retry.
run_atdd "task worker process shutdown" -p mcp-server --bin mcp-server -- --test-threads=1
run_atdd "durable task retry attempt spans" -p crank-runtime --test stages retry -- --test-threads=1
# Story 4.4 (FR-22, FR-23): safe retries, cancellation and outcome_unknown. No new selector is added
# here on purpose: the Story 4.2 lines above already run `crank-core --lib task::` (attempt/retry
# bounds), `crank-registry --test integration tasks::` (checkpoint, resolution, cancel and recovery
# CAS) and `mcp-server --test integration durable_tasks::` (application → PG → worker attempts,
# cancellation through the application entry and the composition oracles). Duplicating one of them
# here would only run the same bounded set twice.