e2e: follow the provider words, folded landings and folded skills - #1680
Merged
Merged
Conversation
Four TestTUIE2E subtests went red on dev 461a43f while the product did what the manual says; each was run alone before and after: - plain_launch_opens_connections_and_harnesses waited for `models` in /connect; since #1513 the group is `providers`. It now filters for the seeded account and requires its connected mark. - a_nested_landing_asks_and_a_key_answers_it and a_refused_landing_is_incomplete pressed their key repeatedly, waiting for a receipt on a card that now folds to one row (#1607). One press answers: they now press once, require the tasks' new state on screen, and wait for the durable receipt in the conversation. - foreign_skills_reach_the_conversation waited for the carried-skill line on a turn that now folds (#1504, #1607). The answers carried all three code words; the driver opens the `▸ worked` disclosure and requires the skill line beside the matching answer. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…swer lands skills-a-turn-used.md said the `skills · …` line stays under the message and is not folded with the turn's steps. Since #1627 (#1607's clean chat) the skill-loading note is retained work and folds into `▸ worked`; the e2e driver in this PR already opens that fold to find it. The page now says what the binary does. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
santoshkumarradha
added
the
area:tests
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
make test-e2e-tuion dev461a43fb6(the 2026-09-29 pre-release pass, before dev moved to staging) passed 15 of 19. The four reds were the suite, not the product — each subtest was run alone before and after this change, and the product did what the manual says every time.plain_launch_opens_connections_and_harnesses/connectgroups model sources underproviderssince #1513, notmodelsa_nested_landing_asks_and_a_key_answers_itaanswers it and recordsyou took this as donea_refused_landing_is_incompletenonce, requiresincomplete, waits for the receiptforeign_skills_reach_the_conversation▸ workeddisclosure and requires the skill line beside the matching answerNo product code or prompt change. One manual correction, and a placement for the owner to confirm: skills-a-turn-used.md said the
skills · …line stays under the message, outside▸ worked. #1627 deliberately changed that (countWork's carried-line exemption became "Skill-loading notes are retained work", and #1607's entry lists skills among what spilled outside live activity), while the same PR's #1504 entry and this page kept the old placement. The page now says what the binary does. If the old placement is the one wanted, it is a three-line restore ininternal/tui3/workfold.goplus reverting this driver hunk and the page paragraph. Needles stay ininternal/e2e/tuiwords_test.go, so the untagged gate still checks them againstinternal/tui3.Proof: each subtest alone after the change: PASS (21.8 s, 20.7 s, 20.7 s, 99.2 s; skills ran both
memory_onandmemory_off).go test ./internal/e2e/(untagged),go vet -tags e2e ./internal/e2e/,make test-laws,make pr-ready: green.🤖 Generated with Claude Code