fix: template contract test passing

[alxiong]

Also closes #3

This PR:

This PR fixed some contract setup so that forge test now all passes

[alxiong]

@akonring you raised a good question, I wasn't too sure about how the Gateway contract works before.

based the doc here, i think we should use Gateway to auto-select the right verifier in practice.

so the vm.mockCall basically says the next call is not actually being executed, just return a mock answer, so the fact that our forge test passes doesn't mean the Plonk proof is being verified. (but simply trivially skipped)

That said, it's tricky to switch to an actual verifier in our test, because we use the fixture.json directly introduced in succinctlabs/sp1-project-template#4, and the first 4 bytes of proof, supposed to be the identifier for the verifier, but we don't control that. None of the existing Verifier contract matches that.
(e.g. the v1.0.7-testnet is expecting a hash of 0x8c5bc5e4 from here, yet the fixture uploaded by succinct team has proof that starts with 0x1b24bf0e)

I also did a quick try to replace the first 4 bytes of the proof in the current fixture.json, but I would get wrong proof size:

[FAIL. Reason: revert: wrong proof size] test_ValidFibonacciProof() (gas: 42733)

So in short, Succint team just uploaded a useless, out of nowhere proof that cannot be verified by any of their existing PlonkVerifier.
This could be fixed once we actually run the 256GB RAM of proof gen, and repopulate the fixture.json.

Maybe leave this PR open, and I will ask Mat for a machine to do that.

[akonring]

That said, it's tricky to switch to an actual verifier in our test, because we use the fixture.json directly introduced in succinctlabs/sp1-project-template#4, and the first 4 bytes of proof, supposed to be the identifier for the verifier, but we don't control that. None of the existing Verifier contract matches that. (e.g. the v1.0.7-testnet is expecting a hash of 0x8c5bc5e4 from here, yet the fixture uploaded by succinct team has proof that starts with 0x1b24bf0e)

I see. Thanks for checking this out.

So in short, Succint team just uploaded a useless, out of nowhere proof that cannot be verified by any of their existing PlonkVerifier. This could be fixed once we actually run the 256GB RAM of proof gen, and repopulate the fixture.json.

Yes, it seems like they only generated the fixture.json for the initial version (v1.0.4-testnet) and "forgot" to generate the the same fixture with subsequent versions (including the versions where the 4-byte verifier check has been introduced).

Maybe leave this PR open, and I will ask Mat for a machine to do that.

Yes, agree. There might be a couple of options/things to keep in mind:

1. We could make as much progress as possible w/o the fixture and just mock contract verification (as we do now).
2. We could ask Succinct Labs to generate the fixture.json for newer versions. This would give us something concrete for a e2e setup with actual verification.
3. We could try to set up a beefy server and generate the proof ourselves using existing SP1 tooling. We'll have to do this eventually so this could give us a head start on potential issues in this process.

[alxiong]

Yes, it seems like they only generated the fixture.json for the initial version (v1.0.4-testnet) and "forgot" to generate the the same fixture with subsequent versions (including the versions where the 4-byte verifier check has been introduced).

amazing, I didn't find their v1.0.4-testnet yesterday, didn't know that it comes from there. Glad you dig into this rabit hole.

We could try to set up a beefy server and generate the proof ourselves using existing SP1 tooling. We'll have to do this eventually so this could give us a head start on potential issues in this process.

I prefer this.

[alxiong]

I spent some time improving on it and got rid of the annoying script/build.rs that's causing the problem.
in 87d8799

now we should merge this soon as it will affect @mrain 's PR.

I will address the outdated fixture.json for contract test in a separate PR