Techcable

> As I'm reading through it, I wonder if there might be a less code-intensive and more readable approach. How about putting some kind of backchannel into the Flush message...

Idea for a simpler implementation: Use a two-way channel or queue dedicated to flushing and acknowledgment of flushing. **EDIT**: A zero-capacity channel may be what we want. According to `crossbeam-channel`:...

> Probably a proper way to do flush ack using only stdlib is to use `CondVar` paired with a `Mutex`. The `u64` acts as a counter. Before flush the sender...

While I think this is a reasonable default for `.typos.toml` files, I think not checking `Cargo.toml` and `pyproject.toml` is a severe footgun. Even worse, there is no way to override...

Both of these changes make sence. A `Log` trait alias and changing `slog::Never` to `std::convert::Infailable` are both good ideas.

> Alternatively, in the `slog` crate itself, it's also possible to `impl slog::Value for LazyLock` directly. Both would require an MSRV of Rust 1.80 (or making it optional behind a...

Another question. Why does [`slog::FnValue`] not do what you need? It should only ever be called once. [`slog::FnValue`]: https://docs.rs/slog/latest/slog/struct.FnValue.html

``` exadgerated -> exaggerated allocing -> allowing, allocating triat -> trait hereustics -> heuristics impercise -> imprecise resiude -> residue holomoprhic -> holomorphic ```

This is more difficult to do than in the `log` crate, because we need to construct an `&'static slog::RecordStruct` at compile time. It turns out that `Location::caller` does not respect...

I spent several hours tracking this down, and it is arguably a bug in the compiler.