e2e: five TestTUIE2E needles rot on clean dev #1014
Description
replication
On a clean origin/dev tree (d906707, #1012):
make build
go test -tags e2e -count=1 -timeout 30m ./internal/e2e/ -run 'TestTUIE2E/(one_figure_on_every_spend_surface|a_nested_landing_asks_and_a_key_answers_it|a_refused_landing_is_incomplete|a_refused_task_proposal_draws_no_schema_sentence|space_in_the_task_room_pages_the_card)'All five FAIL on the clean tree, ~6 min, a few cents. Full suite run (make test-e2e-tui) fails the same five and passes twelve.
needles that never land (20–30s waits each)
one_figure_on_every_spend_surface—spendws(tui_e2e_test.go:1570)a_nested_landing_asks_and_a_key_answers_it—nobody could check it/[a] accept[s] tell it/your call/?(tui_e2e_test.go:255)a_refused_landing_is_incomplete—nobody could check it/[n] not right(tui_e2e_test.go:307)a_refused_task_proposal_draws_no_schema_sentence—1 yes, set it up · 0 no · c change(refusedargs_e2e_test.go:58)space_in_the_task_room_pages_the_card—enter open its room/· waiting in this conversation · alt+athennew conversation(tui_e2e_test.go:1928, :1960)
evidence
Two full runs (branch with unrelated session-economy fixes, and clean baseline) both fail the same five; twelve pass in both. This is the #184 class again — needles rotting against the surface. Filed from a public CI-reachable assertion; needs OPENROUTER_API_KEY or a profile key per liveKey.
acceptance
The five subtests pass on a clean origin/dev tree end to end, either because the needles track the surface again or because the surface is repaired.
Drafted with CodeAF · reviewed and owned by the author
Activity
AbirAbbas commentedon Sep 16, 2026
Closing in favour of #1048, which is the current account of the same reds.
Of the five needles listed here, three pass on
devnow: the two landing needles were respelled by #1004 (merged after the tree this issue measured;[a] accept/[n] not rightbecamea accept/n not right), and the spend-surface subtest was green in #1048's full run on 2026-09-15 ate2ae913b7, where 14 of 17 subtests passed. The two that still fail,a_refused_task_proposal_draws_no_schema_sentenceandspace_in_the_task_room_pages_the_card, are in #1048 with their actual causes (waits on a card foot the model does not always draw) and current acceptance, so that is the issue to work from.