fix: CIVC components share a transcript #14144

iakovenkos · 2025-05-07T17:02:40Z

iakovenkos · 2025-05-16T13:57:23Z

barretenberg/cpp/src/barretenberg/stdlib/goblin_verifier/goblin.test.cpp

- * circuits
- *
- */
-TEST_F(GoblinTests, MultipleCircuits)


It's a duplicate of GoblinRecursiveVerifierTests::NativeVerification

iakovenkos · 2025-05-16T14:04:08Z

barretenberg/cpp/src/barretenberg/stdlib/goblin_verifier/goblin_recursive_verifier.cpp

@@ -14,24 +14,24 @@ namespace bb::stdlib::recursion::honk {
 */
 GoblinRecursiveVerifierOutput GoblinRecursiveVerifier::verify(const GoblinProof& proof)
 {
+    // Verify the final merge step
+    MergeVerifier merge_verifier{ builder, transcript };


the ECCVM challenges used in compute_translation_opening_claims() sub-protocol must depend on the commitments to merge wires

iakovenkos · 2025-05-16T14:11:53Z

barretenberg/cpp/src/barretenberg/stdlib/goblin_verifier/goblin_recursive_verifier.test.cpp

@@ -227,7 +236,24 @@ TEST_F(GoblinRecursiveVerifierTests, TranslatorMergeConsistencyFailure)
        auto [proof, verifier_input] = create_goblin_prover_output();
        // TODO(https://github.com/AztecProtocol/barretenberg/issues/1298):
        // Better recursion testing - create more flexible proof tampering tests.
-        proof.translator_proof[0] += 1; // tamper with part of the limb of op commitment


this is actually modifying accumulated_result, so the CircuitChecker would output false, but for a wrong reason

iakovenkos · 2025-05-16T14:22:13Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

 }

 /**
 * @brief Construct the hiding circuit  then produce a proof of the circuit's correctness with MegaHonk.
 *
 * @return HonkProof - a Mega proof
 */
-std::pair<HonkProof, ClientIVC::MergeProof> ClientIVC::construct_and_prove_hiding_circuit()
+HonkProof ClientIVC::construct_and_prove_hiding_circuit()


cleaner functionality

merge proof has to happen after proving the hiding circuit, as the latter contains the last subtable

iakovenkos · 2025-05-16T14:42:14Z

barretenberg/cpp/src/barretenberg/ultra_honk/decider_verifier.cpp

@@ -21,11 +21,6 @@ DeciderVerifier_<Flavor>::DeciderVerifier_(const std::shared_ptr<DeciderVerifica
    , transcript(transcript)
 {}

-template <typename Flavor>
-DeciderVerifier_<Flavor>::DeciderVerifier_(const std::shared_ptr<DeciderVerificationKey>& accumulator)


redundant constructor

iakovenkos · 2025-05-16T15:48:40Z

barretenberg/cpp/src/barretenberg/stdlib/client_ivc_verifier/client_ivc_recursive_verifier.hpp


  public:
    using Proof = ClientIVC::Proof;
    using FoldVerifierInput = FoldingVerifier::VerifierInput;
    using GoblinVerificationKey = Goblin::VerificationKey;
    using Output = GoblinRecursiveVerifierOutput;

-    struct VerifierInput {


unused struct

iakovenkos · 2025-05-16T15:53:34Z

barretenberg/cpp/src/barretenberg/stdlib/test_utils/tamper_proof.hpp

-    } else {
-        inner_prover.transcript->deserialize_full_transcript();
-    }
+


this tool can't be used with ECCVM, Mega, and MegaZK anymore (cause they don't have serde methods in Flavor::Transcript), so I simplified it a little and wrote a tool below that works for those Flavors

iakovenkos · 2025-05-16T15:56:05Z

barretenberg/cpp/src/barretenberg/stdlib/test_utils/tamper_proof.hpp

-    if constexpr (!std::is_same_v<InnerFlavor, ECCVMFlavor>) {
-        inner_proof = inner_prover.export_proof();
+
+    inner_proof = inner_prover.export_proof();


somewhat ugly because IPA transcript in the flavors with IPAAcc is not a part of the main transcript. decided to keep it less invasive in UltraProver class and introduce such assignments in tests instead

iakovenkos · 2025-05-16T15:58:53Z

...tenberg/cpp/src/barretenberg/stdlib/translator_vm_verifier/translator_recursive_verifier.cpp

@@ -185,6 +185,15 @@ void TranslatorRecursiveVerifier_<Flavor>::verify_consistency_with_final_merge(
        // These are witness commitments sent as part of the proof, so their coordinates are already in reduced form.
        // This approach is preferred over implementing assert_equal for biggroup, as it avoids the need to handle
        // constants within biggroup logic.
+        bool consistency_check_failed = (merge_commitment.y.get_value() != translator_commitment.y.get_value()) ||


as we keep modifying merge, it's probably useful to have this print statement here

barretenberg/cpp/src/barretenberg/stdlib_circuit_builders/mega_flavor.hpp

iakovenkos · 2025-05-16T16:09:43Z

barretenberg/cpp/src/barretenberg/goblin/goblin.cpp

+ *
+ * @return Goblin::MergeProof
+ */
+Goblin::MergeProof Goblin::prove_final_merge()


I may try to keep a single prove_merge() method if this seems unreasonable.

I dont love this but I also dont think it would be any better to have a single method with a default argument or something. It might eventually be the case that the\ merge always needs to share its transcript with something else (e.g. the PG prover in the main protocol) so maybe that will take care of this. I think its ok for now.

…om:AztecProtocol/aztec-packages into si/shared-transcript-mega-eccvm-translator

ledwards2225

Overall looks great. Left a few comments and minor suggestions. If you can get rid of the defaulted transcript input in any of the constructors I think that would be valuable. Its not ideal to have test-only functionality leak into the production constructors

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

ledwards2225 · 2025-05-19T20:10:11Z

barretenberg/cpp/src/barretenberg/eccvm/eccvm_flavor.hpp

@@ -920,538 +920,6 @@ class ECCVMFlavor {

    using VerifierCommitments = VerifierCommitments_<Commitment, VerificationKey>;



Hopefully we're not giving up to much by getting rid of this. There's a world where we could write failure tests that check every single one of these individual components but its certainly a bit brittle as it is now.

barretenberg/cpp/src/barretenberg/eccvm/eccvm_verifier.cpp

ledwards2225 · 2025-05-19T20:17:13Z

barretenberg/cpp/src/barretenberg/eccvm/eccvm_verifier.hpp

+    // Constructor accepting existing transcript
+    explicit ECCVMVerifier(const std::shared_ptr<Transcript>& transcript)
+        : transcript(transcript){};
+
    explicit ECCVMVerifier(const std::shared_ptr<VerificationKey>& verifier_key)


Can these two constructors just go away now that the VK is fixed?

done! also removed the default constructor

ledwards2225 · 2025-05-19T20:20:51Z

barretenberg/cpp/src/barretenberg/goblin/goblin.cpp

+ *
+ * @return Goblin::MergeProof
+ */
+Goblin::MergeProof Goblin::prove_final_merge()


I dont love this but I also dont think it would be any better to have a single method with a default argument or something. It might eventually be the case that the\ merge always needs to share its transcript with something else (e.g. the PG prover in the main protocol) so maybe that will take care of this. I think its ok for now.

barretenberg/cpp/src/barretenberg/stdlib/goblin_verifier/goblin_recursive_verifier.test.cpp

ledwards2225 · 2025-05-19T20:38:06Z

barretenberg/cpp/src/barretenberg/stdlib/honk_verifier/oink_recursive_verifier.hpp

@@ -35,7 +35,7 @@ template <typename Flavor> class OinkRecursiveVerifier_ {
     */
    explicit OinkRecursiveVerifier_(Builder* builder,
                                    const std::shared_ptr<RecursiveDeciderVK>& verification_key,
-                                    std::shared_ptr<Transcript> transcript,
+                                    const std::shared_ptr<Transcript>& transcript,


yay, no default!

...tenberg/cpp/src/barretenberg/stdlib/translator_vm_verifier/translator_recursive_verifier.cpp

barretenberg/cpp/src/barretenberg/ultra_honk/mega_transcript.test.cpp

barretenberg/cpp/src/barretenberg/stdlib_circuit_builders/mega_flavor.hpp

Closes AztecProtocol/barretenberg#1262

single transcript in goblin

15f8aac

iakovenkos self-assigned this May 7, 2025

iakovenkos added 12 commits May 12, 2025 08:35

civc works natively, transcript requires a fix

1055447

Merge branch 'master' into si/shared-transcript-mega-eccvm-translator

1cfb059

Merge branch 'master' into si/shared-transcript-mega-eccvm-translator

7dd690b

debugging bad proof failure failure

5d7c996

civc tests fixed + minor clean-up

8e25d6b

clean up failure and transcript structure tests

4289936

civc failure tests + clean-up

361e765

test fixes

7ecb375

fix transcript tests

1617277

Merge branch 'master' into si/shared-transcript-mega-eccvm-translator

7f716e9

undo unwanted test patterns changes

f4bb010

tests fixed

a358e64

iakovenkos commented May 16, 2025

View reviewed changes

barretenberg/cpp/src/barretenberg/stdlib_circuit_builders/mega_flavor.hpp Show resolved Hide resolved

iakovenkos commented May 16, 2025

View reviewed changes

iakovenkos and others added 4 commits May 16, 2025 16:10

comments, clean-up, add issue

42e926b

fix build

8bdc2d5

Merge branch 'master' into si/shared-transcript-mega-eccvm-translator

72a1807

Merge branch 'si/shared-transcript-mega-eccvm-translator' of github.c…

37b26c0

…om:AztecProtocol/aztec-packages into si/shared-transcript-mega-eccvm-translator

iakovenkos marked this pull request as ready for review May 16, 2025 16:20

iakovenkos requested review from jeanmon, IlyasRidhuan and fcarreiro as code owners May 16, 2025 16:20

iakovenkos requested a review from ledwards2225 May 16, 2025 16:59

iakovenkos changed the title ~~fix: CIVC rec verifier components share a transcript~~ fix: CIVC components share a transcript May 16, 2025

fcarreiro removed their request for review May 16, 2025 17:52

ledwards2225 approved these changes May 19, 2025

View reviewed changes

iakovenkos and others added 2 commits May 20, 2025 13:55

address review comments

4c2a786

Merge branch 'master' into si/shared-transcript-mega-eccvm-translator

bfc8f33

iakovenkos enabled auto-merge May 20, 2025 14:45

iakovenkos added this pull request to the merge queue May 20, 2025

Merged via the queue into master with commit 2ebdc8d May 20, 2025
4 checks passed

iakovenkos deleted the si/shared-transcript-mega-eccvm-translator branch May 20, 2025 15:31

This was referenced May 20, 2025

StructureTest should retrieve the proof from proof_data rather than calling export_proof AztecProtocol/barretenberg#923

Open

ClientIVCRecursiveVerifier should only have one transcript and one proof AztecProtocol/barretenberg#1026

Closed

Thunkar pushed a commit that referenced this pull request May 23, 2025

fix: CIVC components share a transcript (#14144)

7a8942c

Closes AztecProtocol/barretenberg#1262

iakovenkos mentioned this pull request May 28, 2025

Transcript method add_to_hash_buffer() is incompatible with IPA fuzzer AztecProtocol/barretenberg#1408

Open

ledwards2225 mentioned this pull request May 28, 2025

Fix memory fragmentation due to transcript manifest AztecProtocol/barretenberg#1410

Closed

		@@ -920,538 +920,6 @@ class ECCVMFlavor {

		using VerifierCommitments = VerifierCommitments_<Commitment, VerificationKey>;

fix: CIVC components share a transcript #14144

fix: CIVC components share a transcript #14144

Uh oh!

Conversation

iakovenkos commented May 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ledwards2225 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

iakovenkos commented May 7, 2025 •

edited

Loading