François Bobot

> My current concern is that the only sensible semantics of my cyclic example seems to be a dependency cycle error, but the current implementation does something else. I completely...

> Looks like the implementation is almost new, so I'll have to re-review it. Yes the modifications in main were quite big. I'm afraid that we lost the ability to...

Thanks that's good to know. I can remove them, it was only to have that `%{unknown} && false` be directly false instead of building and waiting for `%{unknown}`. But we...

`target[s]` is optional, thanks! > (note: I wish stanzas were listed in alphabetic order in the stanza reference, they are hard to find in the list) Indeed it is always...

I think there is no obligation to add a library in the toplevel scope when it is used by the interpreter.

The simplest would be to modify upstream utop to add the feature through dune rpc and Topcommon.add_directive. We can also build a custom toplevel that adds the feature. Or allows...

We have two meanings: - [opam `install` files](https://opam.ocaml.org/doc/Manual.html#lt-pkgname-gt-install) where `libexec` is the same directory than `lib` but with the bit set. - Distribution layout, where `libexec` is different from `lib`...

So internally I propose to have: | | internally named | opam install file | dune install | |---------|---|----------|-------------| | lib without exec bit (used for .cmo, .a, ...) |...

Partly linked to ocaml/merlin#737 for the use with OCaml tooling. But for C/C++ it has an interest in itself.

I agree with @Khady more generally many more things than just the binary are needed. So If you want `dune exec` to keep up to date the local installation something...