[BUG] Release COM1 editor CI probe fails after catalog and earlier editor checks pass #37

New issue

Open

opened 2026-09-13 00:07:55 +02:00 by erikinkinen · 0 comments

erikinkinen commented

2026-09-13 00:07:55 +02:00

Owner

Summary

The positive release-image command-history and cursor probe over COM1 failed in Integration PR run 1582 at 9e8e2917de855d6bf42285c0f0cfabb116790b26. All 486 catalog scenarios, all four directory-probe routes, development serial time, and the preceding development physical, development COM1 and release physical editor probes passed. The failed release COM1 editor reports serial-input failure with composed VM-command status 125; the available log does not identify the failed edit or establish its cause.

Reproduction

The observed original workflow step was Validate release command history and cursor through COM1:

./tests/native-editor-vm.sh serial out/appliance/erix-release.img

The maintained wrapper selects the workflow's release image and the native-editor profile. Its ten edit/echo actions cover command history, draft restoration, cursor movement, delete/backspace edges, fragmented serial escape input and wrapped long-line editing. This report records one retained CI observation; no replay was performed and frequency is unknown.

Expected and actual behavior

Expected: all ten actions satisfy their exact output, ordered prompt/caret milestones and required pixel checks for the two wrapped lines and final prompt; the positive wrapper returns 0.

Actual: the release COM1 attempt began at 2026-09-12 21:56:59.8468566 UTC. At 21:57:17.4629906 UTC, the sanitized diagnostic was:

vm diagnostic: stage=serial-input runner_exit=1 vm_exit=125 expected_vm_exit=0 physical_input_exit=unknown serial_input_exit=1 physical_cleanup_requested=0

The wrapper then reported failure and returned 1. Main job 19531 failed; Rust and Markdown succeeded. The original runner checks waited serial-input failure before its composed VM-status branches. vm_exit=125 is the composed watchdog/timeout/QEMU-command observation, not an independently recovered raw QEMU exit or proof of a particular watchdog branch. The physical-cleanup field does not establish absence of all cleanup.

Environment and identity

Integration revision: 9e8e2917de855d6bf42285c0f0cfabb116790b26, pull-request run 1582, main job 19531, attempt 1. Its exact component graph is selected by the original components.toml and cargo-repos.toml at that revision.
Image/session: workflow-built release image, maintained native-editor wrapper, serial COM1 input. The failing probe image digest and effective executable byte/version receipts were not independently recovered from these job logs; no artifact download was made.
Declared runner configuration: x86-64 QEMU, one virtual CPU, 120-second VM limit, 15-second progress-stall limit and 10-second command limit. Effective acceleration, firmware and toolchain identities were not inferred from a local environment.
The editor profile expects an 80-column terminal and exact edit/output/caret behavior, including three verified frame observations. These are the source's acceptance requirements, not independently inspected frame evidence from this failed attempt.
No introducing revision or last known working revision for this exact release COM1 observation is established.

Evidence and prior checks

Integration run 1582 is terminal failure. All three terminal job logs were captured completely once: 13,249,625 bytes total. The main log is 13,144,936 bytes, SHA-256 b4b03c517e813a02f62d1607ddd033fe2efaa1dbab25ef96a21a660e7c2f3727.

One-based raw LF anchors in the main log are: 486-case catalog summary 100213; the four directory successes 100771, 100773, 100775 and 100777; development serial-time success 100779; development physical editor success 100781; development COM1 editor success 100783; release physical editor success 100785; failed release COM1 start 100786; diagnostic 100787; wrapper failure 100788. All 486 unique worker pass lines are bound to the original scenario inventory. Nineteen original Git source blobs cover the workflow, wrapper, runner, input and diagnostic contracts. The API records the main job stopped at 2026-09-12 21:58:19 UTC.

The terminal log does not locate a failed edit, output/prompt/caret assertion, framebuffer check, transport operation or cleanup step. Earlier intentional negative host-fixture diagnostics are separate from this positive-probe failure. No host-contention, guest-fault, shared-cause or introducing-change inference is made. No probe, deadline or acceptance gate was changed or rerun.

Further original release observation — 19 September 2026: Integration run 1677, original head 78557a6c, fails the release COM1 editor probe. All 489 catalog scenarios, both native Kernel diagnostics, development physical and COM1 editor checks, and the release physical editor check pass first. Rust and Markdown jobs pass. Three complete hashed job logs total 13,385,246 bytes with zero warning candidates. The main log is 13,280,609 bytes, SHA256 2fea4f1cbfdf3436db7c1703626708f2271b9163da9dd67000cf0654c5b14eac.

At 17:03:29 UTC the input-stage diagnostic reports runner 1, composed VM 125 (expected 0), serial input 1, physical input unknown, no requested physical cleanup and an unavailable last event. The workflow order and preceding successes identify the release COM1 route. The unavailable inner input evidence does not identify a failed edit, frame assertion or root cause. No unchanged rerun, deadline change or original-workflow cancellation occurred. Its companion run 1678 fails earlier at the development COM1 editor route and is separately tracked in issue 67; a common cause remains unproven.

Tracking

Related: Integration #13, the earlier release physical editor observation; and Integration #36, the release COM1 directory observation. This PR passed both the release physical editor and release COM1 directory gates before its later release COM1 editor failure. Their common cause is unproven. Integration PR #12 coordinates the work. No workaround or correction is claimed here.

## Summary The positive release-image command-history and cursor probe over COM1 failed in Integration PR run 1582 at `9e8e2917de855d6bf42285c0f0cfabb116790b26`. All 486 catalog scenarios, all four directory-probe routes, development serial time, and the preceding development physical, development COM1 and release physical editor probes passed. The failed release COM1 editor reports serial-input failure with composed VM-command status 125; the available log does not identify the failed edit or establish its cause. ## Reproduction The observed original workflow step was **Validate release command history and cursor through COM1**: ```sh ./tests/native-editor-vm.sh serial out/appliance/erix-release.img ``` The maintained wrapper selects the workflow's release image and the `native-editor` profile. Its ten edit/echo actions cover command history, draft restoration, cursor movement, delete/backspace edges, fragmented serial escape input and wrapped long-line editing. This report records one retained CI observation; no replay was performed and frequency is unknown. ## Expected and actual behavior Expected: all ten actions satisfy their exact output, ordered prompt/caret milestones and required pixel checks for the two wrapped lines and final prompt; the positive wrapper returns 0. Actual: the release COM1 attempt began at 2026-09-12 21:56:59.8468566 UTC. At 21:57:17.4629906 UTC, the sanitized diagnostic was: ```text vm diagnostic: stage=serial-input runner_exit=1 vm_exit=125 expected_vm_exit=0 physical_input_exit=unknown serial_input_exit=1 physical_cleanup_requested=0 ``` The wrapper then reported failure and returned 1. Main job 19531 failed; Rust and Markdown succeeded. The original runner checks waited serial-input failure before its composed VM-status branches. `vm_exit=125` is the composed watchdog/timeout/QEMU-command observation, not an independently recovered raw QEMU exit or proof of a particular watchdog branch. The physical-cleanup field does not establish absence of all cleanup. ## Environment and identity - Integration revision: `9e8e2917de855d6bf42285c0f0cfabb116790b26`, pull-request run 1582, main job 19531, attempt 1. Its exact component graph is selected by the original `components.toml` and `cargo-repos.toml` at that revision. - Image/session: workflow-built release image, maintained native-editor wrapper, serial COM1 input. The failing probe image digest and effective executable byte/version receipts were not independently recovered from these job logs; no artifact download was made. - Declared runner configuration: x86-64 QEMU, one virtual CPU, 120-second VM limit, 15-second progress-stall limit and 10-second command limit. Effective acceleration, firmware and toolchain identities were not inferred from a local environment. - The editor profile expects an 80-column terminal and exact edit/output/caret behavior, including three verified frame observations. These are the source's acceptance requirements, not independently inspected frame evidence from this failed attempt. - No introducing revision or last known working revision for this exact release COM1 observation is established. ## Evidence and prior checks [Integration run 1582](https://git.erikinkinen.fi/erix/integration/actions/runs/1582) is terminal failure. All three terminal job logs were captured completely once: 13,249,625 bytes total. The main log is 13,144,936 bytes, SHA-256 `b4b03c517e813a02f62d1607ddd033fe2efaa1dbab25ef96a21a660e7c2f3727`. One-based raw LF anchors in the main log are: 486-case catalog summary 100213; the four directory successes 100771, 100773, 100775 and 100777; development serial-time success 100779; development physical editor success 100781; development COM1 editor success 100783; release physical editor success 100785; failed release COM1 start 100786; diagnostic 100787; wrapper failure 100788. All 486 unique worker pass lines are bound to the original scenario inventory. Nineteen original Git source blobs cover the workflow, wrapper, runner, input and diagnostic contracts. The API records the main job stopped at 2026-09-12 21:58:19 UTC. The terminal log does not locate a failed edit, output/prompt/caret assertion, framebuffer check, transport operation or cleanup step. Earlier intentional negative host-fixture diagnostics are separate from this positive-probe failure. No host-contention, guest-fault, shared-cause or introducing-change inference is made. No probe, deadline or acceptance gate was changed or rerun. Further original release observation — 19 September 2026: [Integration run 1677](https://git.erikinkinen.fi/erix/integration/actions/runs/1677), original head [78557a6c](https://git.erikinkinen.fi/erix/integration/commit/78557a6c672ecf426dfe894a01cc4aeec73b5e3c), fails the release COM1 editor probe. All 489 catalog scenarios, both native Kernel diagnostics, development physical and COM1 editor checks, and the release physical editor check pass first. Rust and Markdown jobs pass. Three complete hashed job logs total 13,385,246 bytes with zero warning candidates. The main log is 13,280,609 bytes, SHA256 `2fea4f1cbfdf3436db7c1703626708f2271b9163da9dd67000cf0654c5b14eac`. At 17:03:29 UTC the input-stage diagnostic reports runner 1, composed VM 125 (expected 0), serial input 1, physical input unknown, no requested physical cleanup and an unavailable last event. The workflow order and preceding successes identify the release COM1 route. The unavailable inner input evidence does not identify a failed edit, frame assertion or root cause. No unchanged rerun, deadline change or original-workflow cancellation occurred. Its companion [run 1678](https://git.erikinkinen.fi/erix/integration/actions/runs/1678) fails earlier at the development COM1 editor route and is separately tracked in [issue 67](https://git.erikinkinen.fi/erix/integration/issues/67); a common cause remains unproven. ## Tracking Related: [Integration #13](https://git.erikinkinen.fi/erix/integration/issues/13), the earlier release physical editor observation; and [Integration #36](https://git.erikinkinen.fi/erix/integration/issues/36), the release COM1 **directory** observation. This PR passed both the release physical editor and release COM1 directory gates before its later release COM1 editor failure. Their common cause is unproven. [Integration PR #12](https://git.erikinkinen.fi/erix/integration/pulls/12) coordinates the work. No workaround or correction is claimed here.

erikinkinen added the

bug

phase-6

labels

2026-09-13 00:07:55 +02:00

~~erikinkinen referenced this issue 2026-09-13 00:07:56 +02:00~~

WIP: Coordinate native authority, shell workspace and source transport #12

erikinkinen referenced this issue

2026-09-13 02:30:25 +02:00

[BUG] Release physical native-utility CI probe stalls after earlier input checks pass #38

erikinkinen referenced this issue

2026-09-13 04:33:10 +02:00