fix(ci): make redis_setexz a clean noop when redis is unavailable

[AztecBot]

Problem

The nightly barretenberg debug build failed: https://github.com/AztecProtocol/aztec-claude/actions/runs/27398926124

The job dies almost immediately (~1s of script time) with:

--- Run barretenberg-debug CI ---
gzip: stdout: Broken pipe
##[error]Process completed with exit code 1.

Root cause

ci3/source enables set -euo pipefail (via source_options). The first thing bootstrap_ec2 does is:

echo "CI booting..." | redis_setexz "$CI_LOG_ID" 300   # ci3/bootstrap_ec2:24

and redis_setexz was:

function redis_setexz {
  gzip | redis_cli -x SETEX $1 $2 &>/dev/null
}

redis_cli is intentionally a noop when redis is unavailable (CI_REDIS_AVAILABLE=0) — but a noop that never reads its stdin. So the upstream gzip in the pipe gets SIGPIPE/EPIPE, fails, and under pipefail aborts the whole script before the build can even start. cache_log already guards this exact case by draining stdin (cat >/dev/null) when redis is unavailable; redis_setexz did not, so it remained a latent failure for any environment where redis can't be reached (mirror repos, external contributors, local runs without docker).

In normal aztec-packages CI redis is always reachable, so this path never triggered. It surfaced on the aztec-claude mirror nightly, where AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY / BUILD_INSTANCE_SSH_KEY are empty, so no redis tunnel is established and CI_REDIS_AVAILABLE=0.

Fix

Guard redis_setexz on CI_REDIS_AVAILABLE exactly like redis_cli/cache_log already do, draining stdin when redis is unavailable so the upstream producer never hits a broken pipe.

Verified red/green in isolation under set -euo pipefail with CI_REDIS_AVAILABLE=0: the old function aborts the caller (exit 141 / broken pipe); the new function returns cleanly and the caller continues.

Secondary note (not fixed here — config/sync, not code)

The deeper reason the scheduled nightly runs on aztec-claude at all is that the mirror's next is stale. The workflow in this repo already carries a guard added in 4a284902a0a (2026-06-04, "chore: don't run scheduled jobs on forks"):

if: ${{ github.event_name != 'schedule' || github.repository == 'AztecProtocol/aztec-packages' }}

The aztec-claude copy is pinned to a 2026-05-20 upstream merge — before that guard landed — so it still runs the scheduled job on the mirror. Once aztec-claude syncs next, the scheduled run will be skipped there entirely. This PR fixes the underlying script fragility regardless, so the broken pipe can't recur in any no-redis environment.

---

Created by claudebox · group: slackbot

[AztecBot]

Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed.