[BUG] Factory-failure host fixture surfaces stopped serial publication during cleanup #82

Open
opened 2026-09-24 13:58:40 +02:00 by erikinkinen · 0 comments
Owner

Summary

The transport host test test_factory_failure_still_quits_private_vm errors during fixture cleanup in original CI run 1773, before the VM catalog starts. The cause and first affected revision remain unestablished.

Reproduction

The retained original workflow builds its pinned graph and invokes:

python3 -B -Werror tests/interactive_shell_transport_test.py

The workflow supplies the image and prerequisite setup. Original workflow and relevant fixture/runner Git blobs were checked against the retained revision. No new VM reproduction, unchanged-input retry, deterministic trigger or failure-frequency estimate is claimed.

Expected and actual behavior

A deliberately failing transport factory should still terminate its private fake guest and join every owned worker. Expected publication cancellation during acknowledged shutdown should remain distinguishable from unrelated worker failures, which must not be silently discarded.

The retained traceback goes from guest.join() through the guarded serial worker and initial banner publication to fixture_socket_io.send, which raises WriteStopped: fixture publication interrupted by stop. The fixture joins its worker and propagates this exception. The traceback alone does not establish the exact ordering of the factory exception, QMP quit acknowledgement and serial publication. This is a fake-socket host observation, not an EriX guest failure.

Environment and identity

  • Original run 1773, Integration 7207a703823c0a754674025777f99dbcaeb3db65, API run 10431, integration job 22406, attempt 1.

Host Python fake-socket transport fixture using nonbootable fixture files. The original source and timing limits are unchanged.

Evidence and prior checks

Rust 320/321-test configurations and Markdown pass; the catalog VM suite and canonical image gates are not reached. Run 1773: three complete terminal logs, 258,365 bytes; main log SHA-256 c5d17b6adb8f54f4abf0d73b0e7291b529f7e5a07e677fb5d5982d531a8fce64. Complete retained log lengths and hashes verify, with zero warning candidates. Passing neighboring runs do not repair these failures. The failed-stage observations do not attribute a Kernel regression or establish an authority violation. Related fixture ownership and serial-write reports: #41 and #42.

Tracking

Track diagnosis and correction through Integration PR 12, integration validation and Phase 6 completion. Preserve the original timing, output, consent and cleanup contracts. Keep this report open until an attributed correction passes the unchanged applicable requirements.

## Summary The transport host test `test_factory_failure_still_quits_private_vm` errors during fixture cleanup in original CI run 1773, before the VM catalog starts. The cause and first affected revision remain unestablished. ## Reproduction The retained original workflow builds its pinned graph and invokes: ```sh python3 -B -Werror tests/interactive_shell_transport_test.py ``` The workflow supplies the image and prerequisite setup. Original workflow and relevant fixture/runner Git blobs were checked against the retained revision. No new VM reproduction, unchanged-input retry, deterministic trigger or failure-frequency estimate is claimed. ## Expected and actual behavior A deliberately failing transport factory should still terminate its private fake guest and join every owned worker. Expected publication cancellation during acknowledged shutdown should remain distinguishable from unrelated worker failures, which must not be silently discarded. The retained traceback goes from `guest.join()` through the guarded serial worker and initial banner publication to `fixture_socket_io.send`, which raises `WriteStopped: fixture publication interrupted by stop`. The fixture joins its worker and propagates this exception. The traceback alone does not establish the exact ordering of the factory exception, QMP quit acknowledgement and serial publication. This is a fake-socket host observation, not an EriX guest failure. ## Environment and identity - [Original run 1773](https://git.erikinkinen.fi/erix/integration/actions/runs/1773), Integration `7207a703823c0a754674025777f99dbcaeb3db65`, API run 10431, integration job 22406, attempt 1. Host Python fake-socket transport fixture using nonbootable fixture files. The original source and timing limits are unchanged. ## Evidence and prior checks Rust 320/321-test configurations and Markdown pass; the catalog VM suite and canonical image gates are not reached. Run 1773: three complete terminal logs, 258,365 bytes; main log SHA-256 `c5d17b6adb8f54f4abf0d73b0e7291b529f7e5a07e677fb5d5982d531a8fce64`. Complete retained log lengths and hashes verify, with zero warning candidates. Passing neighboring runs do not repair these failures. The failed-stage observations do not attribute a Kernel regression or establish an authority violation. Related fixture ownership and serial-write reports: #41 and #42. ## Tracking Track diagnosis and correction through [Integration PR 12](https://git.erikinkinen.fi/erix/integration/pulls/12), [integration validation](https://git.erikinkinen.fi/erix/integration/issues/3) and [Phase 6 completion](https://git.erikinkinen.fi/erix/integration/issues/65). Preserve the original timing, output, consent and cleanup contracts. Keep this report open until an attributed correction passes the unchanged applicable requirements.
erikinkinen added this to the Alpha 1 milestone 2026-10-30 09:39:20 +01:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
erix/integration#82
No description provided.