Fix weight generation for pallet-asset-manager

[ahmadkaouk]

What does it do?

This PR fix an issue with weight generation for the pallet-asset-manager caused by changes introduced in #2908.

The functions change_existing_asset_type and remove_existing_asset_type in the WeightInfo trait are not supposed to accept arguments, as indicated by their signatures:

pub trait WeightInfo {
	fn register_foreign_asset() -> Weight;
	fn change_existing_asset_type() -> Weight;
	fn remove_existing_asset_type() -> Weight;
}

But since we're using this syntax le x in 5..100 in the benchmark scenarios, this generates the following code:

    /// Storage: `AssetManager::AssetIdType` (r:1 w:1)
    /// Proof: `AssetManager::AssetIdType` (`max_values`: None, `max_size`: None, mode: `Measured`)
    /// Storage: `AssetManager::AssetTypeId` (r:0 w:2)
    /// Proof: `AssetManager::AssetTypeId` (`max_values`: None, `max_size`: None, mode: `Measured`)
    /// The range of component `x` is `[5, 100]`.
    fn change_existing_asset_type(x: u32, ) -> Weight {
        // Proof Size summary in bytes:
        //  Measured:  `447 + x * (4 ±0)`
        //  Estimated: `3909 + x * (5 ±0)`
        // Minimum execution time: 16_455_000 picoseconds.
        Weight::from_parts(18_175_493, 3909)
            // Standard Error: 1_387
            .saturating_add(Weight::from_parts(77_450, 0).saturating_mul(x.into()))
            .saturating_add(T::DbWeight::get().reads(1_u64))
            .saturating_add(T::DbWeight::get().writes(3_u64))
            .saturating_add(Weight::from_parts(0, 5).saturating_mul(x.into()))
    }
    /// Storage: `AssetManager::AssetIdType` (r:1 w:1)
    /// Proof: `AssetManager::AssetIdType` (`max_values`: None, `max_size`: None, mode: `Measured`)
    /// Storage: `AssetManager::AssetTypeId` (r:0 w:1)
    /// Proof: `AssetManager::AssetTypeId` (`max_values`: None, `max_size`: None, mode: `Measured`)
    /// The range of component `x` is `[5, 100]`.
    fn remove_existing_asset_type(x: u32, ) -> Weight {
        // Proof Size summary in bytes:
        //  Measured:  `447 + x * (4 ±0)`
        //  Estimated: `3909 + x * (5 ±0)`
        // Minimum execution time: 14_125_000 picoseconds.
        Weight::from_parts(15_904_692, 3909)
            // Standard Error: 1_352
            .saturating_add(Weight::from_parts(70_484, 0).saturating_mul(x.into()))
            .saturating_add(T::DbWeight::get().reads(1_u64))
            .saturating_add(T::DbWeight::get().writes(2_u64))
            .saturating_add(Weight::from_parts(0, 5).saturating_mul(x.into()))
    }

Which does not match the signature in the trait.

[github-actions]

WASM runtime size check:

Compared to target branch

Moonbase runtime: 2260 KB (no changes) ✅

Moonbeam runtime: 2248 KB (no changes) ✅

Moonriver runtime: 2240 KB (no changes) ✅

Compared to latest release (runtime-3300)

Moonbase runtime: 2260 KB (+232 KB compared to latest release) ⚠️

Moonbeam runtime: 2248 KB (+252 KB compared to latest release) ⚠️

Moonriver runtime: 2240 KB (+248 KB compared to latest release) ⚠️

[github-actions]

Coverage Report

@@                        Coverage Diff                         @@
##           master   ahmad-fix-weights-asset-manager     +/-   ##
==================================================================
  Coverage   74.58%                            74.58%   0.00%     
  Files         375                               375             
  Lines       95682                             95682             
==================================================================
  Hits        71356                             71356             
  Misses      24326                             24326             

Files Changed	Coverage

Coverage generated Thu Dec 5 11:01:22 UTC 2024

[RomarQ]

@ahmadkaouk can you explain the issue? The CI benchmarks check was passing.

[ahmadkaouk]

@RomarQ I’ve added a description of the issue in the PR. The problem is that the compilation fails after the weights are generated. To address this, we should consider adding an extra step in the CI benchmark check to compile the runtime with the generated weights.

[gonzamontiel]

I'm not sure whether this is the right approach to remove the loop, but it looks in sync with the definition and fix the benchmark.

[RomarQ]

This should be fine since we are no longer using num_assets_weight_hint;