Overview

7 days30 days

Latest CI Pipeline Executions

Fuzzy

Succeeded
35545
22afaea9 fix(core): vendor turso_sdk_kit patch to forward NoLock to WAL open Confirmed via local two-process repro that the patch unlocks multi-process DB opens. Upstream pre.26 has a one-line bug in turso_sdk_kit/src/rsapi.rs: it computes open_flags with NoLock when experimental_multiprocess_wal is enabled, applies it to the initial .db file open, but then hardcodes OpenFlags::default() on the inner Database::open_with_flags_async call — which is what opens the .db-wal file with an exclusive lock that blocks every other process. Vendored copy of turso_sdk_kit-0.6.0-pre.26 lives under cargo-patches/, referenced via [patch.crates-io] in Cargo.toml. The only delta vs the registry source is in src/rsapi.rs around the Opening phase — search for "PATCH:" to find it. Drop this once upstream cuts a release with the fix.
by Jason Jean
Failed
35545
6f299717 chore(repo): retrigger CI on turso multiprocess_wal commit Previous run was canceled before e2e tests started, so we don't know whether experimental_multiprocess_wal(true) actually changed Linux behavior. Empty commit to re-fire the pipeline.
by Jason Jean
Canceled
35545
346d2063 fix(core): enable experimental_multiprocess_wal on turso Builder Without this flag turso 0.6.0-pre.26 silently uses an in-process WAL backend, so a second OS process opening the same DB hits "File is locked by another process" — the same failure mode that killed the original spike (#34933). Note: with the flag set, the lock contention moves from .db to .db-wal but still rejects a second process in real daemon+CLI flows (verified via local two-process repro). Pushing anyway so CI can exercise the new code path on Linux x64 and we can see if the behavior is platform-specific or universally broken.
by Jason Jean
Failed
35545
762451ed feat(core): replace rusqlite with turso 0.6.0-pre.26 (multi-process WAL) Replaces the rusqlite C-bindings backend with the pure-Rust turso crate (v0.6.0-pre.26), which now ships process-shared WAL coordination via the upstream tshm protocol (tursodatabase/turso#6236, merged 2026-04-20). The previous spike (#34933) closed because turso could not share a DB file across the daemon, CLI, and plugin workers. The new release fixes that, so the spike can land using a published crate version instead of a git-branch dependency. Removes the rusqlite-era retry/lock-file/WAL-negotiation/WSL1 detection machinery that existed solely to work around rusqlite concurrency edge cases; turso's busy_timeout + WAL handle the same cases natively. Cargo.toml: rusqlite 0.32.1 -> turso 0.6.0-pre.26 (no C deps, WASM-ready) db/initialize.rs: ~500 -> ~120 lines; open + WAL pragma + create tables db/connection.rs: turso::Connection wrapper with DbValue/DbRow shims db/mod.rs: drop lock-file dance; keep DB_VERSION filename + cleanup cache/cache.rs: port fetch_cache_rows from rarray to dynamic IN clause tasks/task_history.rs, details.rs, running_tasks_service.rs: DbValue params telemetry/mod.rs, service.rs: DbValue params, DbRow results Adds initialize_db_concurrent_connections test verifying two connections can simultaneously open and read/write the same DB file.
by Jason Jean