General multi-request signing

[Antonio95]

This PR adds an example test with the general external-signing flow for arbitrary transactions (that is, transactions with any call structure and record flows).

• The test case is “Should match ExecutionRequest.sign for arbitrary flow with multiple requests” in sdk/tests/external-signing.test.ts. It is specifically run on a call to ldgbatcher_p28/transfer_private_3, but the flow is independent of that transaction’s structure.
• An important piece of this PR relies on snarkVM’s sample_authorization_extended from Extend sample_authorization to facilitate multi-request flows snarkVM#3310, so this PR is in draft mode and has a path-patched Cargo.toml until the snarkVM PR is merged. The snarkVM changes are now meged and the snarkVM dependencies in the SDK have been bumped to a revision containing those changes.
• The test matches the delegated-signing flow of our signing spec (which is amply commented in the test itself):
  • Authorizer has view_key but no private key.
  • Signer has the private key.
  • Authorizer calls sample_authorization_extended. Then it engages in one round of communication per request (i.e. per call in the transaction) with the signer as follows:
    1. Authorizer preprocess the request’s inputs for delegated signing, computing the to_fields representation for most types and (H, tag) for static-record inputs.
    2. Signer completes input processing and signs the requests, sending back the signature, the tvk and the gammas of the static-record inputs.
    3. Authorizer computes the input ID and puts together a complete, correct request.
    4. Authorizer uses the record-tracking information (from sample_authorization_extended) to update the nonces of record-like (static/external/dynamic-record) inputs to subsequent requests.
• In order to run the above flow, this PR implements (or exposes from snarkVM) a couple of extra structures and methods:
  • sampleAuthorizationExtended (exposed from snarkVM), corresponding to the step before 1 above.
  • fromPreprocessedInputs, corresponding to step 2 above and returning a new structure PreprocessedSigningResult. It internally calls the also new signRequestFromPreprocessedInputs.
  • computeMintedNonce and applyMintedNonce, corresponding to step 4 above.
  • authorizeRequests (exposed from snarkVM), which is called after step 4 in the test. This is optional, and I believe in our case it is done on the the DPS side.

Tests

• As indicated above, only one new test case is left on the PR, “Should match ExecutionRequest.sign for arbitrary flow with multiple requests”. This case performs the aforementioned flow and makes a few sanity checks on the result.
• It was important to check the resulting requests can be proved and verified, but (as far as I can tell) the SDK repo is ill-equipped for that. So, in commit 8263ed4, I (read: the agent) added the necessary machinery to be able to locally mimic a ledger, prove, and verify the transaction. It worked (and it helped identify a bug, after fixing which the verification passed). The machinery is complex and I don’t think we want to keep it around (also, the test became long to run), so it was removed in the next commit and only a couple of relevant small pieces were added back in the commit after.
• Even though only one transaction is tested here (although one with an already fairly interesting record flow), the snarkVM tests from Extend sample_authorization to facilitate multi-request flows snarkVM#3310 check sample_authorization_extended returns the correct mocked requests and (importantly) record-tracking machinery on hundreds of transactions, including quite wild flows of static/dynamic/external records and conversion between them — so we can be confident on that front.

---

Note

High Risk
Changes authorization and request signing in WASM plus a breaking network-client API; incorrect nonce patching or preprocessed signing would produce invalid authorizations.

Overview
Adds delegated signing for arbitrary multi-call transactions: authorizer holds a view key, signer holds the private key, and requests are signed one transition at a time with nonce patching when minted records flow into later calls.

WASM / snarkVM (bumped to 8e2397ee…) exposes sampleAuthorizationWithRecordTracking (mock auth plus record-tracking, record names, and program checksums), authorizeRequests (build an Authorization from already-signed ExecutionRequests), ExecutionRequest.fromPreprocessedInputs / signRequestFromPreprocessedInputs (sign from computeExternalSigningInputs output), and Group.gScalarMultiply.

SDK adds computeMintedNonce (on-chain cast nonce from TVK + output index) and exports it from the browser bundle. getProgramAmendmentCount is a breaking change: it now requires an edition argument and uses /program/.../amendment_count?edition=; ProgramManager.resolveEditionAndAmendment first calls getLatestProgramEdition.

Tests add a long end-to-end case on ldgbatcher_p28.aleo/transfer_private_3 (four requests, join/transfer chain) and update program-import mocks for the two-step edition/amendment lookup.

^{Reviewed by Cursor Bugbot for commit 9e9276b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Pull request overview

Copilot reviewed 9 out of 10 changed files in this pull request and generated 6 comments.

[Antonio95]

It's true, this boils down to the mix of representations here: https://github.com/ProvableHQ/sdk/blob/mainnet/wasm/src/programs/request.rs#L759-L762. I will fix as suggested (i.e. maintain both and be able to handle them) since I don't know whether changing to "functionId" will have any unexpected consequences (e.g. other libraries deserializing this objects breaking).

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 81266c7. Configure here.}

[iamalwaysuncomfortable]

Some tests were failing + lints were failing, I fixed those.

Otherwise this PR is very well done, very excellent first contribution to the SDK!

[Antonio95]

@iamalwaysuncomfortable I've pushed a couple of commits to try and fix the last remaining CI errors. Can you please confirm this change a24bf5e makes sense?

Only two CI failures remain:

• create-leo-app loyalty-program (delegated + scanner)
• Transport-aware proving (testnet). Here I've fixed an API call (cf. 9e9276b) and the test passes locally, but it fails on the remote CI flow.