fix(e2e): era-aware --reuse-values pin assertion — baseline release now carries the prod pin

[LukasWodka]

Summary

Unblocks #454 and every other chart-touching PR: the "Fleet auto-upgrade E2E (k3d)" check has failed since yesterday's develop → main promotion (#383), because the path-1 assertion hardcoded an era that promotion ended.

Cause.--reuse-values replays the baseline release's computed values and ignores new chart defaults. The assertion expected jm_ingestor_digest to be empty after that upgrade, which was only true while the last published release predated images.ingestor.prodDigest (#398). The #383 promotion published a release that carries the pin in its computed values, so the replay now (correctly) delivers it, and the hardcoded expectation fails — exactly the era boundary the test's own comment described. Diagnosis first posted on #454.

Fix. The expectation is now derived from the baseline release itself, right after it is installed:

• BASELINE_PROD_DIGEST="$(helm get values --all -o json | jq -r '.images.ingestor.prodDigest // ""')"
• Pin absent in baseline ⇒ assert it does not arrive (the old behavior, still correct against pre-fix(chart): move the prod ingestor pin into chart defaults so it reaches edges #398 releases).
• Pin present ⇒ assert the same digest is replayed verbatim — strictly stronger than "non-empty": it still proves the documented limitation (nothing injected from the new chart's defaults) in the new era.

Era-stale comments updated (header contract, path 1, path 2). Other assertions untouched.

Flagged, not fixed here (path-2 era note added in code). Path 1's --reuse-values rewrites the release's recorded values to the baseline's full computed set, so path 2's --reset-then-reuse-values replays the baseline pin as if user-supplied. Today baseline pin == working-tree pin, so path 2's GOT == WANT holds — but the first PR that bumpsprodDigest will fail path 2. That failure is a genuine signal, not noise: any real edge that was ever hand-upgraded with --reuse-values stops receiving chart pin updates the same way. Worth a follow-up decision (isolate path 2 from path 1, or treat as the intended tripwire for the replay-contamination issue).

Type

• Fix (CI)

Test plan

• bash -n + shellcheck clean.
• Era logic: baseline without pin ⇒ old assertion path (unchanged behavior); baseline with pin ⇒ verbatim-replay assertion. jq is preinstalled on the stock runners.
• The real proof is this PR's own Helm Chart CI run: its E2E installs the current (post-Release develop → main #383, pin-carrying) release and exercises the new era-B branch end-to-end.

🤖 Generated with Claude Code

---

Note

Low Risk
CI-only changes to e2e-auto-upgrade.sh assertions and comments; no production Helm chart or runtime behavior is modified.

Overview
Fixes the Fleet auto-upgrade E2E (k3d) gate after published charts started carrying images.ingestor.prodDigest in computed values.

The path-1 --reuse-values check no longer assumes the prod ingestor pin is always absent. After installing the last published release, the script reads BASELINE_PROD_DIGEST from helm get values --all (via jq, with a new preflight) and branches: no baseline pin ⇒ digest must stay empty; baseline pin present ⇒ digest must match the baseline verbatim (proving new chart defaults are not injected on this path).

Header and inline comments are updated to describe both eras. Path 2 behavior is unchanged; a new ERA NOTE documents that path 1 can contaminate stored values so a future prodDigest bump may fail path 2 on purpose (client#459).

^{Reviewed by Cursor Bugbot for commit 5e582dc. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 8879713. Configure here.}

[saqlainsyed007]

A minor Suggestion:

jq as a hard dependency for local runs. The claim "jq is preinstalled on the stock runners" is true for GitHub-hosted Ubuntu, but the script advertises local use (Usage: bash scripts/tests/e2e-auto-upgrade.sh) and already gates docker via the has helper. A one-line has jq || error "jq is required" in the preflight section would fail with a clear message instead of a mid-run pipeline abort.

[LukasWodka]

Good catch — done in 3d13b3d: has jq || error "jq is required (it reads the baseline release's computed values)." sits right next to the docker gate in the preflight section, so a local run without jq fails immediately with a clear message instead of a mid-run pipeline abort. Cherry-picked to #454 as well (d76af0a) so the two copies of the fix stay identical for either merge order.

bugbot run

[LukasWodka]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 5e582dc. Configure here.}