While validating PR #66's Responses header configuration change on feature/issue-65-open-responses at HEAD 12dc469 with uncommitted Responses/rig changes, cargo test --workspace --all-features failed three tests in unchanged serdes-ai-cli/tests/ui_collapse.rs:
the_answer_is_still_visible_after_collapsing, line 166: the block did not expand.ctrl_o_with_nothing_shortened_says_so, line 206: timed out after 30s waiting for Nothing to expand; captured screen ended with a literal ^O after the completed turn.expanding_more_than_fits_says_how_much_is_shown, line 196: timed out waiting for the expansion message.
The first run reported 9 passed and 3 failed in this binary. A subsequent run of the same full-workspace command passed all 1,803 tests, with 128 ignored doctests, without any CLI source changes. The cause is not established; this records an intermittent failure rather than attributing it to the Responses change.
Environment: macOS on Apple Silicon, stable Rust toolchain. Local evidence in /Volumes/XS1000/acoliver/projects/serdesAI/branch-1/tmp/pr66-astra-auth/workspace-tests.log (failed run) and workspace-tests-final.log (passing run). These are local artifacts, not repository files. No CLI fix is included in PR #66's header work.
Suggested investigation: check PTY input readiness around the completion-to-input transition and the collapse test's Ctrl-O injection. Repeated runs should establish whether input can arrive before the terminal is ready for it.
While validating PR #66's Responses header configuration change on
feature/issue-65-open-responsesat HEAD12dc469with uncommitted Responses/rig changes,cargo test --workspace --all-featuresfailed three tests in unchangedserdes-ai-cli/tests/ui_collapse.rs:the_answer_is_still_visible_after_collapsing, line 166:the block did not expand.ctrl_o_with_nothing_shortened_says_so, line 206: timed out after 30s waiting forNothing to expand; captured screen ended with a literal^Oafter the completed turn.expanding_more_than_fits_says_how_much_is_shown, line 196: timed out waiting for the expansion message.The first run reported 9 passed and 3 failed in this binary. A subsequent run of the same full-workspace command passed all 1,803 tests, with 128 ignored doctests, without any CLI source changes. The cause is not established; this records an intermittent failure rather than attributing it to the Responses change.
Environment: macOS on Apple Silicon, stable Rust toolchain. Local evidence in
/Volumes/XS1000/acoliver/projects/serdesAI/branch-1/tmp/pr66-astra-auth/workspace-tests.log(failed run) andworkspace-tests-final.log(passing run). These are local artifacts, not repository files. No CLI fix is included in PR #66's header work.Suggested investigation: check PTY input readiness around the completion-to-input transition and the collapse test's Ctrl-O injection. Repeated runs should establish whether input can arrive before the terminal is ready for it.