Summary
testmd run replays a previously-cached plan for a given test even after the source *_test.md steps change, and --author does not reliably force fresh authoring. A test whose first run produced a degenerate 5-step plan keeps returning reason: "replay completed" with that same trivial plan on every subsequent run.
Severity
S2
kane-cli version
0.4.0
Repro
testmd run x_test.md ... (first run authors some plan).
- Edit the steps in
x_test.md.
testmd run x_test.md --author ... and/or rm -rf output-* ~/.testmuai/kaneai/tests/* then re-run.
Actual
Subsequent runs still emit reason: "replay completed" with the old plan; edited steps are not re-authored. Clearing local output-* / ~/.testmuai did not help (plan appears to persist server-side / in TMS).
Expected
--author should bypass the cache and re-author from the current source; or there should be a documented way to invalidate a test's cached plan.
Summary
testmd runreplays a previously-cached plan for a given test even after the source*_test.mdsteps change, and--authordoes not reliably force fresh authoring. A test whose first run produced a degenerate 5-step plan keeps returningreason: "replay completed"with that same trivial plan on every subsequent run.Severity
S2
kane-cli version
0.4.0
Repro
testmd run x_test.md ...(first run authors some plan).x_test.md.testmd run x_test.md --author ...and/orrm -rf output-* ~/.testmuai/kaneai/tests/*then re-run.Actual
Subsequent runs still emit
reason: "replay completed"with the old plan; edited steps are not re-authored. Clearing localoutput-*/~/.testmuaidid not help (plan appears to persist server-side / in TMS).Expected
--authorshould bypass the cache and re-author from the current source; or there should be a documented way to invalidate a test's cached plan.