Skip to content

Windows CI: mcp_boot incremental-boot tests flake on 10s node-fixture deadline #5980

Description

@Hmbown

Flake

mcp_boot_does_not_restore_servers_removed_during_handshake and mcp_boot_reports_ready_server_before_stalled_server_finishes (both via assert_incremental_mcp_boot in crates/tui/src/core/engine/tests.rs) panic on the windows leg with:

fast server must be visible before slow server is released: Elapsed(())

at tests.rs:20951. Seen on #5972 and #5946 windows legs (e.g. job 101614678303); three sibling trees passed windows the same day.

Root cause

The test spawns two node stdio fixture servers and gives the "fast" one 10s to complete its handshake while the "slow" one is stalled on a release file. On a loaded Windows runner, node cold-start + stdio handshake can exceed 10s, so the progress timeout fires. Slow-runner timing-deadline flake, not a regression — same family #5970 fixed elsewhere (wait_for_captured_output, clipboard 3s→30s).

Fix

Give the helper an honest budget (30s for both the progress and finished waits). The contract is ordering (fast visible before slow released), not speed, so a longer deadline does not weaken any assertion.

Audit: codewhale-ops/AUDIT-REPORT-V0913-20260907.md.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    • Status
      Backlog

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions