CI flake: Build smoketests (Windows) #6384

Activity

1. clockwork-labs-bot commented on Oct 26, 2026

   

   clockwork-labs-bot

   on Oct 26, 2026

   ContributorAuthor

   More actions

   Exact-attempt diagnosis

   Inspected only run 40307278855, job 121522038130, attempt 1, Build smoketests (Windows). The job API confirms attempt 1, matching the ticket's failed attempt. This was a pull-request run for #6376, head e6577621e2586cdeb536ee95bce5082431ddcf77.

   • Runner: ghr-56a1a89e, label spacetimedb-windows-runner. The job started at 10:31:24 UTC and was marked failed at 10:42:25 UTC on October 26.
   • Setup, checkout, Rust toolchain setup, OpenSSL/cache setup, and cargo-nextest installation succeeded. Build smoketest dependencies and archive test binaries started at 10:31:39 UTC and remains recorded as in_progress; subsequent statistics, timing/artifact uploads, and post-checkout steps remain pending.
   • The exact job's logs endpoint returns HTTP 404. Its check-run failure annotation explicitly says: “The self-hosted runner lost communication with the server.” The ticket's log-fetch error is a reporting symptom, not the initiating failure.
   • Root cause established by this attempt: Windows self-hosted runner/server communication loss during build work, a CI infrastructure failure rather than an evidenced compiler or test failure. The annotation does not establish whether the deeper cause was host/runner termination, resource starvation, or network interruption. Runner/host telemetry is needed to distinguish those possibilities; no compiler diagnostic or test assertion is available. This is not evidence of a build-code failure or a test hang. No other attempt was inspected for this diagnosis.

   Suggested updated title: CI flake: Windows self-hosted runner lost communication during Build smoketests

   Duplicate analysis

   Searched public issues across open and closed states using broad failure-mechanism terms runner lost and Windows logs, then inspected matching Test flakes tickets and their discussions/duplicate state.

   Top duplicate candidate: #5799, the open canonical Windows runner-loss/no-log tracker. Its consolidation explicitly includes Windows build jobs with unavailable logs.

   • CI flake: Build smoketests (Windows) #6297 is a close historical repeat in Test flakes: the same build job, HTTP 404 logs, runner communication-loss annotation, and interrupted build step. It is already closed as a duplicate, with its diagnosis pointing to CI flake - Windows self-hosted runner loses communication with GitHub #5799; the canonical candidate is therefore CI flake - Windows self-hosted runner loses communication with GitHub #5799, not CI flake: Build smoketests (Windows) #6297.
   • CI flake - Windows runner loses communication with GitHub during setup #5821 is another related Test flakes manifestation with the same communication-loss annotation and missing logs, but interrupted during dependency setup rather than build work; its followup also identifies CI flake - Windows self-hosted runner loses communication with GitHub #5799 as canonical.
   • CI flake: Build smoketests (Windows) #5801 is the duplicate-closed Windows build/no-log manifestation explicitly consolidated into CI flake - Windows self-hosted runner loses communication with GitHub #5799.

   The mechanism match is sufficiently close to mark this ticket as a duplicate of #5799 using GitHub duplicate state, not completed state. The unavailable logs limit deeper root-cause attribution, not the observed runner-loss classification.
2. clockwork-labs-bot

   closed this as a duplicateof CI flake - Windows self-hosted runner loses communication with GitHub #5799 on Oct 26, 2026