Automated release: version and notes generated from pull requests merged since 4.1.0.
- A guessed option name is answered with the right one. Every tool now adds to its usage error, for each
--flagit does not have: its nearest option (--treshold: did you mean--threshold?), the options whose help uses the word (straighten.py --angle:--degreesmentions 'angle';fit.py --ratio:--aspect), and the tools that do take a flag of that name (export.py --width:fit.pyand others). It is said even when a required option is also missing, where argparse named only that one.overlay.py --positionlists its names when one is wrong. - First-try stumbles are recorded and replayed.
docs/stumbles.jsonlholds each call an agent got wrong (what it typed, what came back, what was right, the class); a line markedtaughtis run by the tests and the tool's answer must name the way out. Five stumbles where the result is wrong and says nothing are listed asopen. - Option names stop growing apart. The names for an angle, a strength, a text size, a duration and the rest are decided (docs/design-decisions.md, "one name per meaning"). Nothing is renamed while the freeze holds; a test keeps a new option from taking an older name and a shared name from meaning two things. Twenty-four more wrong calls, each another tool's name for the same thing, are in the ledger and answered.
- fix: a guessed option name is answered with the right one, and first-try stumbles are recorded and replayed (#642)
What's Changed
- fix: a guessed option name is answered with the right one, and first-try stumbles are recorded and replayed by @kajisho5 in #642
Full Changelog: v4.1.0...v4.1.1