How do lake compute dependencies usability?

Stream: lean4

[Alissa Tung]

I am building a binary cache for Lean, my questions are:

1. How do lake compute dependencies usability?

I tried the following steps to prepare some cache for a project (target project):

1. clone all the repo that manifest refers to
2. checkout the REV that manifest refers to
3. elan override each deps project with the lean-toolchain of original project (target project)
4. lake build for each deps

The cache step is that, if (URL + REV + lean-toolchain) match, I hope cache would work if I put build artifacts to the .lake/packages.

but when all deps are ready, moved to .lake/packages of target project, in target project exec lake build, after a long time it recompiles from the first module, all of previous artifacts are ignored.

How do lake compute dependencies usability or what method should I use to setup and prepare a binary cache for Lean?