feat: Sync from noir

[AztecBot]

Automated pull of development from the noir programming language, a dependency of Aztec.
BEGIN_COMMIT_OVERRIDE
fix: update in_contract flag before handling function metadata in elaborator (noir-lang/noir#5292)
fix: fix incorrect return type being applied to stdlib functions modulus_be_bytes(), modulus_be_bits(), etc. (noir-lang/noir#5278)
chore: refactor test case generation in build.rs (noir-lang/noir#5280)
fix: handle struct with nested arrays in oracle return values (noir-lang/noir#5244)
feat: build simple dictionary from inspecting ACIR program (noir-lang/noir#5264)
fix: Add more thorough check for whether a type is valid when passing it from constrained code to unconstrained code (noir-lang/noir#5009)
chore: Pedersen commitment in Noir (noir-lang/noir#5221)
chore: pedersen hash in Noir (noir-lang/noir#5217)
fix: Don't lazily elaborate functions (noir-lang/noir#5282)
fix: avoid unnecessarily splitting expressions with multiplication terms with a shared term (noir-lang/noir#5291)
fix: avoid duplicating constant arrays (noir-lang/noir#5287)
feat: add fuzzer for Noir programs (noir-lang/noir#5251)
feat: Run comptime code from annotations on a type definition (noir-lang/noir#5256)
feat: implement comptime support for as_slice builtin (noir-lang/noir#5276)
chore: create separate crate just for noir artifacts (noir-lang/noir#5162)
feat: add support for wildcard types (noir-lang/noir#5275)
chore: replace logical operators with bitwise in DebugToString (noir-lang/noir#5236)
fix: use proper serialization in AbiValue (noir-lang/noir#5270)
chore: simplify compilation flow to write to file immediately (noir-lang/noir#5265)
feat: implement comptime support for array_len builtin (noir-lang/noir#5272)
chore: Use the elaborator by default (noir-lang/noir#5246)
chore: Release Noir(0.31.0) (noir-lang/noir#5166)
feat!: remove dep:: prefix (noir-lang/noir#4946)
feat: Sync from aztec-packages (noir-lang/noir#5242)
chore: replace is_bn254 implementation to not rely on truncation of literals (noir-lang/noir#5247)
chore: add no-predicate to hash implementations (noir-lang/noir#5253)
feat(experimental): Implement macro calls & splicing into Expr values (noir-lang/noir#5203)
feat: add BoundedVec::map (noir-lang/noir#5250)
chore: add no predicate to poseidon2 (noir-lang/noir#5252)
feat: add set and set_unchecked methods to Vec and BoundedVec (noir-lang/noir#5241)
fix: Disable if optimization (noir-lang/noir#5240)
chore: redo typo PR by dropbigfish (noir-lang/noir#5234)
chore: add property tests for ABI encoding (noir-lang/noir#5216)
chore: thread generics through ACIR/brillig gen (noir-lang/noir#5120)
chore: copy across typo PR script from aztec-packages (noir-lang/noir#5235)
chore(docs): fixing trailing slash issue (noir-lang/noir#5233)
fix: add support for nested arrays returned by oracles (noir-lang/noir#5132)
chore: Parse macros (noir-lang/noir#5229)
chore: Optimize the elaborator (noir-lang/noir#5230)
fix(elaborator): Fix regression introduced by lazy-global changes (noir-lang/noir#5223)
fix(elaborator): Fix duplicate methods error (noir-lang/noir#5225)
chore: fixing all relative paths (noir-lang/noir#5220)
chore: push code related to ABI gen into noirc_driver (noir-lang/noir#5218)
END_COMMIT_OVERRIDE

[github-actions]

Changes to circuit sizes

Generated at commit: 2ead89a47ad98753494b07f5ee15c5d686cb67a7, compared to commit: 33ccf1b61260868e6bb027b7838ef530717bff01

🧾 Summary (100% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
rollup_base	+393,777 ❌	+116.12%	+21,006,387 ❌	+678.33%
private_kernel_reset	+299,292 ❌	+123.64%	+17,169,149 ❌	+674.72%
private_kernel_reset_big	+149,532 ❌	+83.60%	+8,584,389 ❌	+504.10%
public_kernel_tail	+516,811 ❌	+42.47%	+27,090,896 ❌	+499.18%
private_kernel_reset_medium	+74,652 ❌	+50.69%	+4,292,009 ❌	+334.78%
private_kernel_reset_small	+37,212 ❌	+28.30%	+2,145,819 ❌	+200.24%
private_kernel_tail	+11,999 ❌	+54.61%	+484,639 ❌	+36.25%
public_kernel_app_logic	+24,261 ❌	+7.12%	+463,098 ❌	+22.86%
public_kernel_teardown	+24,285 ❌	+7.13%	+463,140 ❌	+22.86%
private_kernel_tail_to_public	+11,944 ❌	+2.60%	+482,811 ❌	+22.79%
public_kernel_setup	+11,373 ❌	+4.05%	+306,968 ❌	+17.03%
rollup_root	+2,023 ❌	+94.31%	+144,483 ❌	+15.86%
private_kernel_inner	+5,793 ❌	+10.83%	+120,721 ❌	+9.34%
private_kernel_init	+498 ❌	+1.62%	+25,488 ❌	+6.52%
parity_root	+99 ❌	+4.63%	+7,411 ❌	+0.66%
parity_base	+102 ❌	+29.48%	-28,854 ✅	-36.15%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
rollup_base	732,876 (+393,777)	+116.12%	24,103,174 (+21,006,387)	+678.33%
private_kernel_reset	541,352 (+299,292)	+123.64%	19,713,792 (+17,169,149)	+674.72%
private_kernel_reset_big	328,405 (+149,532)	+83.60%	10,287,293 (+8,584,389)	+504.10%
public_kernel_tail	1,733,658 (+516,811)	+42.47%	32,517,991 (+27,090,896)	+499.18%
private_kernel_reset_medium	221,933 (+74,652)	+50.69%	5,574,045 (+4,292,009)	+334.78%
private_kernel_reset_small	168,696 (+37,212)	+28.30%	3,217,420 (+2,145,819)	+200.24%
private_kernel_tail	33,971 (+11,999)	+54.61%	1,821,451 (+484,639)	+36.25%
public_kernel_app_logic	364,783 (+24,261)	+7.12%	2,489,280 (+463,098)	+22.86%
public_kernel_teardown	365,005 (+24,285)	+7.13%	2,489,561 (+463,140)	+22.86%
private_kernel_tail_to_public	470,511 (+11,944)	+2.60%	2,600,970 (+482,811)	+22.79%
public_kernel_setup	292,333 (+11,373)	+4.05%	2,109,666 (+306,968)	+17.03%
rollup_root	4,168 (+2,023)	+94.31%	1,055,264 (+144,483)	+15.86%
private_kernel_inner	59,304 (+5,793)	+10.83%	1,413,370 (+120,721)	+9.34%
private_kernel_init	31,185 (+498)	+1.62%	416,264 (+25,488)	+6.52%
parity_root	2,238 (+99)	+4.63%	1,136,706 (+7,411)	+0.66%
parity_base	448 (+102)	+29.48%	50,957 (-28,854)	-36.15%

[benesjan]

I tried fixing the remaining issues and the problem is that the resulting contract artifacts now have null as a return_type and we try to map it to our TS interfaces which do not support that.

I see that for other contracts and for functions which actually return something the return_type is not null so this looks like a conscious change and we should just update our TS interface.

@TomAFrench can you please confirm that null return type is expected? Once I have the confirmation I'll update the types.

[TomAFrench]

Yes this is expected in the case where we don't have a return value as we can't assign it a type.

[TomAFrench]

The only thing that makes sense a potential cause for this issue is that we have an issue in how the new pedersen hash implementation deals with predicates. This could potentially screw up the calculation of the args_hash in the below code.

aztec-packages/noir-projects/aztec-nr/aztec/src/hash.nr

Lines 111 to 136 in fd92d46

	pub fn hash_args(args: [Field]) -> Field {
	if args.len() == 0 {
	0
	} else {
	assert(args.len() < ARGS_HASH_CHUNK_COUNT * ARGS_HASH_CHUNK_LENGTH);
	let mut chunks_hashes = [0; ARGS_HASH_CHUNK_COUNT];
	let mut current_chunk_values = [0; ARGS_HASH_CHUNK_LENGTH];
	let mut current_chunk_index = 0;
	let mut index_inside_current_chunk = 0;
	for i in 0..args.len() {
	current_chunk_values[index_inside_current_chunk] = args[i];
	index_inside_current_chunk+=1;
	if index_inside_current_chunk == ARGS_HASH_CHUNK_LENGTH {
	chunks_hashes[current_chunk_index] = pedersen_hash(current_chunk_values, GENERATOR_INDEX__FUNCTION_ARGS);
	current_chunk_values = [0; ARGS_HASH_CHUNK_LENGTH];
	current_chunk_index+=1;
	index_inside_current_chunk = 0;
	}
	}
	if index_inside_current_chunk > 0 {
	chunks_hashes[current_chunk_index] = pedersen_hash(current_chunk_values, GENERATOR_INDEX__FUNCTION_ARGS);
	}
	pedersen_hash(chunks_hashes, GENERATOR_INDEX__FUNCTION_ARGS)
	}
	}

[sirasistant]

One question, are the generators the same in noir's implementation of pedersen vs BB implementation of pedersen? we might be having a mismatch between typescript computed pedersen hashes vs noir computed pedersens

[TomAFrench]

Yeah, the issue seems to be that the separator was being applied to the length generator rather than the generators for the commitment.