ivokub
commented
2 weeks ago This is super cool!
I see that some of the public inputs are zero. For efficiency, we by default use the in-complete formulas when computing a multi-scalar multiplication and zero as a public input would hit the edge case.
To force using complete formulas, you can provide the groth16.WithCompleteArithmetic() verification option. In your example you should do:
@@ -416,5 +430,5 @@ func (c *VerifyCircomProofCircuit) Define(api frontend.API) error {
if err != nil {
return fmt.Errorf("new verifier: %w", err)
}
- return verifier.AssertProof(c.VerifyingKey, c.Proof, c.PublicInputs)
+ return verifier.AssertProof(c.VerifyingKey, c.Proof, c.PublicInputs, stdgroth16.WithCompleteArithmetic())
}By the way - we also have the "test engine" which does not create a proof, but instead evaluates the circuit using *big.Int and checks the in-circuit assertions. This is allows for faster debugging of the circuit without needing to set up the keys etc. For example:
@@ -13,6 +13,7 @@ import (
"github.com/consensys/gnark/frontend"
"github.com/consensys/gnark/frontend/cs/r1cs"
"github.com/consensys/gnark/logger"
+ "github.com/consensys/gnark/test"
"github.com/rs/zerolog"
"github.com/consensys/gnark/std/algebra/emulated/sw_bn254"
@@ -229,6 +230,19 @@ func main() {
var pk groth16.ProvingKey
var vk groth16.VerifyingKey
+ // Create the circuit assignment with actual values
+ circuitAssignment := &VerifyCircomProofCircuit{
+ Proof: recursionProof,
+ VerifyingKey: recursionVk,
+ PublicInputs: recursionPublicInputs,
+ }
+
+ err = test.IsSolved(placeholderCircuit, circuitAssignment, ecc.BN254.ScalarField())
+ if err != nil {
+ panic(err)
+ }
+ return
+And it seems that your example works.
p4u
Hello. I managed to create a Parser for Circom/SnarkJS proof, verifying key and inputs to Gnark. I do believe this is a nice to have feature because there is nothing like Circom for browsers (in terms of cmpatibility, memory consumption and so on). The idea behind this is to be able to aggregate Circom proofs using Gnark and the bn254 recursion.
The current status is that the Parser works and I can verify the Circom Proof using the groth16 Verifier from Gnark. That's great!
However, I'm struggling trying to build a second SNARK to prove recursively the Circom Proof (I have been following the example on std/recursion/groth16). So at this point I'm blocked and so asking for help.
Everything seems to work fine until the proof is generated, then after some time computing, it fails:
12:30:22 ERR error="no modular inverse" nbConstraints=3510919.Since the non-recursive verification of the proof works, I do believe the issue is when formatting the types to the new recursion types. I have tried several ways to do it, but the result is always the same.
Any hint that helps me resolve this issue would be very welcome.
The code is messy and dirty, I'll package it correctly if I manage to make it work :)
https://github.com/p4u/circomgnark
(this log output is from the branch
with_circuit_definition)