Move the precompiles to a new directory that is shipped to npm

[PaulRBerg]

As discussed here, it would be helpful to ship the precompiles so that we can write a protocol deployment script that spans V2 Core and V2 Periphery.

Task:

• Move the precompiles in the precompiles directory
• Include the precompiles in the npm package
• Include the script/Precompiles.s.sol file in the npm package (via files)

Food for thought: or maybe we shouldn't include the precompiles in script? How about another directory, e.g., precompiles?

I will let you @andreivladbrg and @smol-ninja self-organize and assign this.

[andreivladbrg]

or maybe we shouldn't include the precompiles in script? How about another directory, e.g., precompiles?

interesting idea, i like it, i've got just one concern, would foundry compile the contracts from a not known path? or does foundry compile contract only from the configured paths? not sure if it is possible to override test:

[profile.default]
  src = "src"
  test = "test"
  test ="precompiles"

[smol-ninja]

Looks like by setting allow_paths, we can instruct foundry to read solidity files from additional directories.

[andreivladbrg]

oh, ok, thanks

thought that allow_paths is only for remappings

[PaulRBerg]

AFAIK, Foundry will be able to read the contracts even without allow_paths defined. Anyway - something that should be easy to test, @andreivladbrg.

[smol-ninja]

If we haven't yet decided on the directory for precompiles, how about using src/precompiles for that?

My Rationale behind it

1. Having it in script appears to be similar to having it in test/utils as both of them are not directly related to the source code. While script stores deployment scripts, test/utils stores utility functions for testing.
2. All the imports in precompile are from src so it looks reasonable to put it in the src directly.

[PaulRBerg]

That's an interesting idea, @smol-ninja.

One wrinkle with putting it in src is that it would break the topology of the current subdirectories there.

That is:

• abstracts
• interfaces
• libraries
• types

precompiles would not be at the same "logical level" with any of the foregoing categories. The precompiles encompass everything in src.

Thus, I would be slightly more inclined to put the precompiles in a new folder called precompiles that sits at the same level as src.

[smol-ninja]

Thanks so much for explaining it @PaulRBerg. After reading your comment, I am now in favour of moving it to precompiles folder.

[smol-ninja]

Clarifications:

1. We need script/Precompiles.s.sol script in v2-periphery and not in v2-core, right?
2. To include Precompiles.s.sol script in the npm package, It should inherit from Script and not BaseScript otherwise we would also have to include Base.s.sol in the npm package, right? Or do we want to include Base.s.sol?

AFAIK, Foundry will be able to read the contracts even without allow_paths defined.

That's correct.

[PaulRBerg]

Sorry @smol-ninja, I had missed your questions here.

I confess that I don't understand your questions.

1. The fact that I've been away from the code base for many months is telling. Why do we need script/Precompiles.s.sol?
2. Precompiles.sol does NOT inherit from either Script or BaseScript.

Could you please re-explain? Your questions might have referred to an older state of the code base.

---

More context: I am looking at the staging branches in V2 Core and V2 Periphery and the way the precompiles have been refactored looks great to me.

[smol-ninja]

No problem @PaulRBerg. I am happy to explain it again.

The fact that I've been away from the code base for many months is telling. Why do we need script/Precompiles.s.sol?

Well it was in the OP (I struckthrough it). Iirc, the idea was to ship it with npm package so that developers can use it to deploy the entire protocol using precompiles.

Precompiles.sol does NOT inherit from either Script or BaseScript.

Yes thats correct. The question was about Precompiles.s.sol.

[PaulRBerg]

Ooh, I see.

I think what I forgot is that Precompiles.s.sol is a hypothetical script that we have never written. Then, you made this remark that settled the matter insofar we realized that a protocol-level deployment script isn't feasible from a gas POV.