Rudi Grinberg

Could you add a test case for this?

I don't know if there's much we can do. We need to load the rules in a directory to see if the executable exists. That's probably where the overhead comes...

It wouldn't be a valid optimization. The rules can change between runs and the stale artifact would remain.

It would not. Re-loading the rules will clear stale artifacts.

I'm going to close as this is just a general performance issue of rule loading and nothing specific to `dune exec`.

Thanks. The warnings looks right, however I think we're not yet ready for this code because we still use opam's library to unpack local archives - but not downloaded ones....

ping @anmonteiro

I don't see any harm with this. In fact, we have something similar internally. Feel free to send a PR

As discussed, the performance penalty currently is too large to build this by default. A separate alias sounds like a good start to me. When this feature becomes fast enough...

> Additionally, is there some way to list the targets built for some given aliases ? And to have Dune output time spent running commands (ideally total time per tool)...