e2e wrapper does not forward IT_MONERO_VIEW_KEY, IT_TARI_VIEW_KEY or IT_TARI_SPEND_PUBLIC_KEY to the detached harness, so payout-confirm SKIPs on the bench

[pithead-developer]

Line refs are against develop @ 71543ca6. Filed from p2pool-starter-stack/bench-ci#668.

Defect. tests/integration/e2e.sh launches the live harness on the bench over ssh (run_harness, the on_bench "... nohup setsid ./.e2e-run.sh ..." launch). The remote command sets only IT_RIG_TOKEN, IT_RIG_ROLLBACK_CHANGES, IT_RIG_POOLS_PROBE, RIG_LOCK_PARENT_* and RIG_LOCK_WAIT. Nothing forwards IT_MONERO_VIEW_KEY, IT_TARI_VIEW_KEY or IT_TARI_SPEND_PUBLIC_KEY. resolve_overrides in tests/integration/lib.sh (around line 194) reads exactly those three to run local-pruned-main-payout-confirm. So a value set in the wrapper's environment never reaches the harness, and the row SKIPs with needs IT_MONERO_VIEW_KEY whatever the caller supplies. The REMOTE_* inputs avoid this because they cross as --remote-* flags. The view keys have no such path. The wrapper's usage (ENV OVERRIDES:) does not list them either.

Evidence. bench-ci job 515@474e306ae3ae14d265604716cd525b739fb558bc (mode: matrix) logged SKIPPED scenario local-pruned-main-payout-confirm: needs IT_MONERO_VIEW_KEY. bench-ci#668 has the operator install the three values in the bench's tier4-e2e environment. Until the wrapper forwards them, the row stays skipped on the bench anyway.

Expected. The wrapper carries the three values to the detached harness the same way it carries IT_RIG_ROLLBACK_CHANGES and IT_RIG_POOLS_PROBE: on stdin, never on the remote command line, where ps on the bench would show them. It lists them under ENV OVERRIDES. A selftest or wrapper test proves that a value set for the wrapper arrives in the harness's environment, and that the row stays skipped when the values are unset.

Related: #2657 (its tier-4 proof is this row), #2497 item 3, #2454.

[pithead-developer]

Blocked on p2pool-starter-stack/bench-ci#678. Tier 4 is still unproven: job 1055@7586bfb2 ended in error before checkout, so my branch never ran on the bench. The bench's chain-sync probe was refused (curl exit 7) until the wait ran out after 2933s. The job has one log line and no artifacts. bench-ci#678 already tracks this failure for every tier4-e2e job from 1041 to 1073, which includes mine, so I'm blocked on that issue.

Everything else on draft PR #2683 is finished: the wrapper change, the selftest (6 of its 26 assertions fail without the fix), and the doc rows. GitHub CI at 7586bfb has 24 checks passing, 2 skipped and none failing. The PR body now records the job's outcome.

Left to do: once bench-ci#678 closes, resubmit tier4-e2e with scenario: local-pruned-main-payout-confirm at this head. That run should show the row running instead of skipping.

The runner resumes this issue when that one closes.

[pithead-developer]

Blocked on p2pool-starter-stack/bench-ci#683. bench-ci#678 is closed, but its fix hasn't reached the benches yet, so I haven't resubmitted. The fix merged as 54ed2577 (via #682), and GET /benches still shows every bench on b82405b6, one commit behind it. I filed bench-ci#683 ("Deploy pending: 54ed2577…") and am blocked on it.

Draft PR #2683 is otherwise complete, and GitHub CI at 7586bfb is green (24 passing, 2 skipped). Job 1055@7586bfb2 errored before checkout because of the bench fault, so tier 4 is still unproven. Next step once the benches report the fix: resubmit tier4-e2e with scenario: local-pruned-main-payout-confirm at 7586bfb.

The runner resumes this issue when that one closes.