Summary

runtime-overrides-e2e fails on every nightly run since being wired into the nightly pipeline by PR #2607 (merged Apr 28). The test builds the Docker image (nemoclaw-override-test) successfully but crashes with exit code 1 at the very first step — "Capturing baseline config (no overrides)" — before any actual override scenarios execute.

No log artifact is produced (the glob test-runtime-overrides-*.log matches nothing), making root-cause analysis from CI alone difficult.

Reproduction

Observed in 3 consecutive nightly runs on main:

• Run 25084401781 (schedule, Apr 29 00:15 UTC)
• Run 25082763579 (dispatch, Apr 28 23:21 UTC)
• Run 25079089844 (dispatch, Apr 28 21:39 UTC)

Error

TEST: Building test image: nemoclaw-override-test
sha256:769410b1d89ebf919038416025ead501f1723e109111ed7555d9772409b95a22
TEST: Capturing baseline config (no overrides)
##[error]Process completed with exit code 1.

The test calls run_override (a docker run --rm that cats /sandbox/.openclaw/openclaw.json) with no env overrides. This suggests the sandbox image either:

1. Does not have /sandbox/.openclaw/openclaw.json at the expected path, or
2. The entrypoint fails before the config file is written, or
3. A Docker/environment assumption (e.g. overlay mounts, user namespace) does not hold on ubuntu-latest GitHub runners.

Related

• PR fix(ci): wire 6 unwired E2E scripts into nightly pipeline #2607 — wired this test into nightly-e2e.yaml
• test/e2e/test-runtime-overrides.sh — the test script

Expected Behavior

The baseline config capture should succeed and the test should proceed through all override scenarios.

Suggested Fix

1. Reproduce locally with docker run --rm nemoclaw-override-test bash -c "cat /sandbox/.openclaw/openclaw.json" to isolate whether it is a build or runtime issue.
2. Add set -x or explicit error logging before the run_override call to capture the Docker stderr.
3. Ensure the log artifact glob matches the actual log file path.