Mark Van de Vyver

@andylokandy, I'm finding more of these corner cases.... some very subtle inter-plays with rustfmt, clippy macros, etc. Given I'm in the midst of migrating the proc-macro code to the new...

> PS: If I didn't get it wrong, the pipeline refactoring could fix those problems, right? It might... depends on how subtle the edge cases are. Some others I see...

> How long will the pipeline refactoring take? Rust's limited integration test support is proving the sticking point. I'm on architecture-attempt #3 to see if I can make it work....

I think this raises an important question. Does minitrace strive for performance first or compatibility with the Log and Tracing ecosystem first? There are three ways to wear the cost...

@jon-chuang and @Xuanwo would you be happy to close this issue, or is your proposal limited to a check at compile time and no runtime checks?

Not sure I understand: > Thus, If `enter_on_poll` becomes the default behavior, the consistency will be broken: To clarify: You'd like there to be consistency, by default, in the number...

> > To clarify: You'd like there to be consistency, by default, in the number of spans generated by one call of a sync function and and one call of...

> > What is the use case/scenario where having async functions appear once adds value? > > The first request that I know so far is [#64 (comment)](https://github.com/tikv/minitrace-rust/issues/64#issuecomment-936271952) Thanks. That...

> Better, in my mind, is to stick with a simple span.record("key", "value") API. In fact we could make the public API even smaller with `record!(span,"key", "value")`

Hopefully we can keep it ***very*** simple - the `tracing` crate world is crate spaghetti. From my PoV. - Reporters just plugin and work (hopefully my `#[trace]` macro work will...