Skip to content

P0: Fix flaky SQLite WAL livenessTest — replace Date() comparison with monotonic clock, remove CI skip #98

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
totalslacker merged 1 commit into from
May 7, 2026
Merged

P0: Fix flaky SQLite WAL livenessTest — replace Date() comparison with monotonic clock, remove CI skip #98

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 5 additions & 20 deletions Tests/SwitchcraftTests/SQLiteStorageConcurrencyTests.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,24 +66,9 @@ struct SQLiteStorageConcurrencyTests {
/// Under the old single-actor design the write queues behind the read and
/// finishes only after the read completes, so `writeEnd ≥ readEnd`.
/// With the writer/reader split, the write completes concurrently so
/// `writeEnd < readEnd`.
///
/// **Disabled on CI runners.** The assertion is `writeEnd < readEnd`
/// based on `Date()` resolution. In release mode on GitHub Actions
/// runners the seeded FTS scan completes fast enough that the read and
/// write timestamps land within the same `Date` tick, making the
/// strict-less-than comparison fail by a tie even when the underlying
/// concurrency is correct. `safariUnfuckerRegressionTest()` continues
/// to gate the same regression with a stronger assertion (~50 sequential
/// writes must not stall against a slow scan), so CI coverage of the
/// writer/reader split is preserved.
@Test(
"Write completes while slow FTS read is in flight (WAL liveness)",
.disabled(
if: ProcessInfo.processInfo.environment["CI"] != nil,
"Date()-resolution comparison too tight for release-mode CI; run locally to validate"
)
)
/// `writeEnd < readEnd`. Uses `ContinuousClock` (nanosecond resolution)
/// so the assertion is reliable even when the FTS scan is fast.
Comment on lines +69 to +70
@Test("Write completes while slow FTS read is in flight (WAL liveness)")
func livenessTest() async throws {
let (storage, url) = try await Self.makeTemporaryDB()
defer {
Expand All @@ -103,11 +88,11 @@ struct SQLiteStorageConcurrencyTests {

// Issue a write on the writer actor and record when it finishes.
try await storage.upsertDocument(Self.makeWriteDoc(uuid: "liveness-write"))
let writeEnd = Date()
let writeEnd = ContinuousClock.now

// Wait for the read to finish and record its completion time.
_ = try await slowRead
let readEnd = Date()
let readEnd = ContinuousClock.now

// The write must finish before the read completes: writer and reader
// run on separate actors/connections. Under the old single-actor design
Expand Down
Loading