ci(dev): make SDK publish fail loudly, decouple ci-base rebuild #72

Merged
will.anderson merged 1 commits from hotfix/ci-dev-publish-hardening into dev 2026-07-15 16:34:20 +00:00
Owner

The dev PUSH build failed (run 3298) while nothing published to foundation-dev.

Root-cause analysis:

  • Publish step had NO set -e — an auth/upload failure exited 0 (silent no-publish).
  • ci-base rebuild runs set -euo pipefail + Docker (pull/build/push ~600MB) on the host-mode GCE runner, where Docker availability is fragile — that is the step that reddened the job.
  • Verified empirically: the CI SA (neuron-ci-pusher) CAN write generic artifacts to foundation-dev and read/push neuron-ci, so this is NOT a missing IAM grant. el_runtime.js IS present on dev, so the .js upload is valid.

Fix:

  1. Publish step: add set -euo pipefail + empty-key guard + active-account echo so publish failures surface with a retrievable log and redden the job (publishing is the deliverable).
  2. ci-base rebuild: continue-on-error: true — it is a CI-cache optimization, not the release artifact, and must never block the publish.

Merging this to dev and re-running the push build will either publish a fresh el-runtime-* version today, or fail loudly at the exact gcloud line for precise diagnosis.

The dev PUSH build failed (run 3298) while nothing published to foundation-dev. Root-cause analysis: - Publish step had NO `set -e` — an auth/upload failure exited 0 (silent no-publish). - ci-base rebuild runs `set -euo pipefail` + Docker (pull/build/push ~600MB) on the host-mode GCE runner, where Docker availability is fragile — that is the step that reddened the job. - Verified empirically: the CI SA (neuron-ci-pusher) CAN write generic artifacts to foundation-dev and read/push neuron-ci, so this is NOT a missing IAM grant. el_runtime.js IS present on dev, so the .js upload is valid. Fix: 1. Publish step: add `set -euo pipefail` + empty-key guard + active-account echo so publish failures surface with a retrievable log and redden the job (publishing is the deliverable). 2. ci-base rebuild: `continue-on-error: true` — it is a CI-cache optimization, not the release artifact, and must never block the publish. Merging this to dev and re-running the push build will either publish a fresh el-runtime-* version today, or fail loudly at the exact gcloud line for precise diagnosis.
will.anderson added 1 commit 2026-07-15 16:33:54 +00:00
ci(dev): make SDK publish fail loudly, decouple ci-base rebuild
El SDK CI - dev / build-and-test (pull_request) Successful in 8m51s
4696fd6833
The dev push build went green-then-red while nothing published: the
Publish step had no set -e, so an auth/upload failure exited 0 (silent
no-publish), while the ci-base rebuild (set -euo pipefail + Docker on the
host-mode runner) hard-failed the job. Add set -euo pipefail + an empty-key
guard + active-account echo to the Publish step so failures surface with a
retrievable log, and mark the ci-base cache rebuild continue-on-error so
the fragile Docker step can never block the actual SDK artifact publish.
will.anderson merged commit 155a449c4e into dev 2026-07-15 16:34:20 +00:00
Sign in to join this conversation.
No Reviewers
No labels
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: neuron-technologies/el#72