start.bats "the tmux session does not carry iw's internal plumbing" fails at baseline: IW_HOME missing from the captured pane env #556

Closed
opened 2026-09-14 10:13:02 +00:00 by mph · 2 comments
Owner

Found while working IW-517.

test/start.bats case "start: the tmux session does not carry iw's internal plumbing" fails on this machine at baseline. Confirmed against 4605fc9c in a clean worktree, so it is unrelated to the IW-517 change.

SYMPTOM: IW_HOME= is missing from the captured pane environment, so the assertion that the session carries no internal plumbing reads a pane that does not carry the variable the test expects to inspect.

Distinct from #468: that issue is about tmux BATS tests failing NON-deterministically under a full suite run. This one fails deterministically, on its own, at baseline.

Needs its own investigation.

Found while working IW-517. `test/start.bats` case "start: the tmux session does not carry iw's internal plumbing" fails on this machine at baseline. Confirmed against `4605fc9c` in a clean worktree, so it is unrelated to the IW-517 change. SYMPTOM: `IW_HOME=` is missing from the captured pane environment, so the assertion that the session carries no internal plumbing reads a pane that does not carry the variable the test expects to inspect. Distinct from #468: that issue is about tmux BATS tests failing NON-deterministically under a full suite run. This one fails deterministically, on its own, at baseline. Needs its own investigation.
Member

Intake: needs-human
Workflow: diagnostic
Reason: G7 — the proposed fix turns every sessionEnv entry into a tmux -e argument, which freezes the 13 secrets your tmux config drops on purpose (a rotated token stays broken); the scope of -e is the one choice a human must make, and the analysis makes it silently
Analysis: project-management/issues/IW-556/analysis.md on branch IW-556

Intake: needs-human Workflow: diagnostic Reason: G7 — the proposed fix turns every sessionEnv entry into a tmux `-e` argument, which freezes the 13 secrets your tmux config drops on purpose (a rotated token stays broken); the scope of `-e` is the one choice a human must make, and the analysis makes it silently Analysis: project-management/issues/IW-556/analysis.md on branch IW-556
Member

Reproduced again while working IW-566, on a different branch and a later baseline.

The same case fails on a clean detached worktree at f3a938bc — a commit that contains none of the IW-566 --no-wait code — and fails identically after the change. IW_HOME= never appears in the captured pane environment in either state.

So the failure is still deterministic, still at baseline, and still unrelated to whatever change is in flight. It makes ./iw ./test non-green on every branch, which costs every phase session a manual "is this mine?" check.

Reproduced again while working IW-566, on a different branch and a later baseline. The same case fails on a clean detached worktree at `f3a938bc` — a commit that contains none of the IW-566 `--no-wait` code — and fails identically after the change. `IW_HOME=` never appears in the captured pane environment in either state. So the failure is still deterministic, still at baseline, and still unrelated to whatever change is in flight. It makes `./iw ./test` non-green on every branch, which costs every phase session a manual "is this mine?" check.
mph closed this issue 2026-09-16 15:21:24 +00:00
Sign in to join this conversation.
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
iterative-works/iw-cli#556
No description provided.