# Nushell core team meeting 2024-08-21
## Attendees
- Jack
- Darren
- Michael
- Yash
- NotTheDroids
- Stefan
- Devyn
- Levertion
- Andy

## Agenda
- [x] Release rollup
- [x] [13662](https://github.com/nushell/nushell/issues/13662) Justin identified a record parsing issue (not sure if it's new)
- [x] PR Input (not super important, just need input)
- [x] Looking for suggestions for better function names for [13633](https://github.com/nushell/nushell/pull/13633)
- [x] IntoValue/FromValue [13641](https://github.com/nushell/nushell/pull/13641)
- [x] FromValue attribute [13647](https://github.com/nushell/nushell/pull/13647)
- [x] Record values with colons [13413](https://github.com/nushell/nushell/pull/13413)
- [x] Set current dir at startup [12953](https://github.com/nushell/nushell/pull/12953) - is this worth one of us fixing the conflicts?
- [x] testing
- being "adverserial" to the implementation
- make it easier to add more cases
- infra improvements

## Discussed Topics

## Release 0.97
- Ultra quick fix of the `..` issue (thanks Devyn and Yash)
- [13633](https://github.com/nushell/nushell/pull/13633) could also be a potentially older issue
- This is seems to be down to quoting and unquoting in processing for external and then in the `nu` binary argument parsing
- issue reacting to the new error message for bare word in assignment rhs
- maybe rephrase for "use quotes, or if you want to execute caret"
- doesn't warrant a point release
## PR party

### [YizhePKU cwd change](https://github.com/nushell/nushell/pull/12953)
- the reedline integration needs explicit work
- could uncover some additional bugs
- author AWOL, devyn could take care

### Piepmatz's derive macro magic

- `bytes` dependency should be fine
- Allow plugin authors to use a serde-like API

### consolidating the naming of the flags and args accross the codebaase

- named flag confusion
- + args/positional
- `Call(Ext)`

## Testing
- Bug fixes without regression tests
- PRs often don't have associated tests
- be more biased towards tests
- let's set up a meeting to discuss a new testing approach (builder?)
- make tests easier
- too many different ways to run tests
- plenty of locations
- (test modalities, tests with config/without, isolation/sandboxing etc.)

### complicated config loading matrix
- [Darren's gist](https://gist.github.com/fdncred/b87b784f04984dc31a150baed9ad2447/)
- very relevant for testing to make sure we have the behavior users will see (or be tolerant to the user test systems)
- potential divergence between Rust internal defaults and the default config
- not having to parse is much faster (also something desirable for tests)

### example tests

- are doctests (not for edgecase testing of command behavior)
- we have too many `None` in result as running has indeterminate output/is dangerous
- some could just be parsed if they are dangerous to run
- test for pure success instead of fixed result