test(op-precompiles): Check subset of l1 precompiles in op

[emhane]

Ref #2189

Checks that subset of l1 precompiles are part of op precompiles wrt hardfork

[codspeed-hq]

CodSpeed Performance Report

Merging #2204 will degrade performances by 2.07%

_{Comparing emhane:precompile-subsets (2643c9f) with main (8ed1b16)}

Summary

❌ 1 regressions
✅ 7 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	precompile bench | ecrecover precompile	197.1 µs	201.2 µs	-2.07%

[emhane]

@rakita could you pls explain how tests pass locally without default features, but not in ci?

[rakita]

Need to check to be sure, but bls does not have a no_std variant so one precompile will be missing.

[rakita]

Need to check to be sure, but bls does not have a no_std variant so one precompile will be missing.

Lets wait for #2210, it should make that test pass.

blst are not supported for no_std but for kona they accelare all precompiles so they are fine without it.

[rakita]

#2249 is done so if we merge this, diff would be same. if there is no bls feature and this precompile is called by default it will throw Fatal error.

[emhane]

found the issue here, when blst is not enabled, we still add all the addresses to Prague with bls12_381_precompiles_not_supported, but not to isthmus

[emhane]

odd , this still works locally, but in ci the output is different

[emhane]

ah it's the precompile function pointers that are different in ci for Prague and isthmus sets...no easy way around that unfortunately. equality of function pointers is needed to see that granite overwrites the bn128 pair. will simplify the difference check to just check for address, to make it pass in ci. that granite overwrites bn128 pair is checked in another test already.

[rakita]

lgtm

[rakita]

Testing that at least the address of the precompile is present is a good quick sanity test.