Surfaced while adopting obsidian-e2e 0.8.2 (e18855d). These failures are pre-existing and environmental, not related to the dep bump: they reproduce byte-identically on a 0.8.1 baseline (git stash + frozen install) in a fresh isolated runner instance.
- tests/e2e/multi-value.test.ts: all 7 tests time out waiting for
.suggestion-item / .metaEditPromptInput - the prompt/suggester modal UI never renders under the headless/unfocused instance.
- tests/e2e/audit-modals.test.ts PROMPT-02: same
.metaEditPromptInput timeout.
- Side effect: their stuck prompt modals cascade modalCount-leak assertion failures into later files in full-suite runs (native-properties, fluid-create) - those files pass 7/7 in a fresh instance, so the cascades are order-dependent pollution, not real failures.
Full data in the 0.8.2 adoption run: 0.8.1 baseline 94 passed | 8 failed vs post-change 89 passed | 13 failed, with the delta fully explained by the cascade class above.
Same family as chhoumann/quickadd#1519 (settings pane renders zero rows under CLI-driven GUI automation) - GUI surfaces not mounting in background instances.
Surfaced while adopting obsidian-e2e 0.8.2 (e18855d). These failures are pre-existing and environmental, not related to the dep bump: they reproduce byte-identically on a 0.8.1 baseline (git stash + frozen install) in a fresh isolated runner instance.
.suggestion-item/.metaEditPromptInput- the prompt/suggester modal UI never renders under the headless/unfocused instance..metaEditPromptInputtimeout.Full data in the 0.8.2 adoption run: 0.8.1 baseline 94 passed | 8 failed vs post-change 89 passed | 13 failed, with the delta fully explained by the cascade class above.
Same family as chhoumann/quickadd#1519 (settings pane renders zero rows under CLI-driven GUI automation) - GUI surfaces not mounting in background instances.