Matthias Beyer

Hi. Thanks for filing this. We should introduce a `cargo-deny` CI check for things like this. The simplest thing is to disable the `ini` backend I guess. This can be...

Hi! If someone contributes it, sure. Problem is, though, that KDL (or XML fwiw) are not tree data formats, which is essentially what config-rs supports only. So it is really...

Nice. Could you also add one with `yaml-rust`, so that we have all three cases? I will then take these patches and continue based on them... not sure how, but...

Yes, I meant add another case where you use _only_ `yaml-rust`, without `config-rs` in between. This way we can verify that we don't mess things up in the `config-rs` implementation!

So if I understand correctly: * Plain `yaml-rust` succeeds * `config-rs` (which uses `yaml-rust`) fails Okay, so we have to investigate why that is.

> The demand for Dhall in `config-rs` seems low. Perhaps this isn't worth landing into `config-rs`? Yes, depending on an unmaintained implementation is definitively something we shouldn't rely on, sorry.

That's indeed an issue, yes. Thanks for raising this! I am happy with either, a patch to fix the issue you raised here, or a patch(set) to move from `nom`...

I do not even have full access myself, so for that we would need to talk to @mehcode ... I hope they respond. In any case, if @epage wants to...

You can use `Config::try_deserialize()` to deserialize config to your own type and have that implement `Serialize`!

Hm, I don't know to be honest. One thing I could imagine is that serializing does more than just "dumping the data to some format", but actually includes logic that...