Add library prefix to all ckzg symbols

[michaelsproul]

Currently the c-kzg library uses #define constants for the mainnet/minimal paramaters like FIELD_ELEMENTS_PER_BLOB:

c-kzg-4844/src/c_kzg_4844.h

Lines 39 to 43 in cfdd9e5

	/*
	* This value represents the number of field elements in a blob. It must be
	* supplied externally. It is expected to be 4096 for mainnet and 4 for minimal.
	*/
	#ifndef FIELD_ELEMENTS_PER_BLOB

This makes it very hard (but not impossible*) to link two copies of the library to enable switching between mainnet and minimal configurations at runtime.

*My attempt at doing this is documented here: sigp/lighthouse#4276. I had to use objcopy to rename c-kzg's symbols so that they wouldn't conflict. Needless to say this is not a sensible approach to take.

To do this better I was thinking the library could accept another #define constant which it prepends to all symbol names. Then the downstream consumer of the library could easily avoid collisions, with names like ckzg_blob_to_kzg_commitment and ckzgmin_blob_to_kzg_commitment instead of just blob_to_kzg_commitment. The ckzg prefix could be the default if no prefix is specified.

[asn-d6]

Hello, that seems like a promising approach to solving the problem, and requires less tinkering of the audited code than making FIELD_ELEMENTS_PER_BLOB configurable. I can look next week at how to achieve this (since honestly I have no idea). Thanks for the suggestion.

[divagant-martian]

digging this approach 🎉

[asn-d6]

So, I pushed #313 which includes some cursed macro code 2c58b2a that gets used around the codebase 6253751 .

To use this cursed feature you need to define a LIB_PREFIX as follows: LIB_PREFIX=ckzg

Would this work for you guys?

Either way, I think we should probably wait until Justin is back from OOO to see if this is acceptable in terms of code quality. I don't really like it, but if it's a needed feature I can probably turn a blind eye.

[michaelsproul]

@asn-d6 That looks like it would work for us, and I don't think the macros are too bad from an aesthetics/maintainability PoV. If I get some time I'll try to re-work the Rust bindings and confirm that it works end-to-end.

Thanks for your work on that!

[ppopth]

Either way, I think we should probably wait until Justin is back from OOO to see if this is acceptable in terms of code quality. I don't really like it, but if it's a needed feature I can probably turn a blind eye.

Yeah, let's wait for his comment, but I think the PR already looks good.

[jtraglia]

Either way, I think we should probably wait until Justin is back from OOO to see if this is acceptable in terms of code quality. I don't really like it, but if it's a needed feature I can probably turn a blind eye.

Yeah, let's wait for his comment, but I think the PR already looks good.

Reviewed the PR yesterday. I'm not ecstatic about it either, but I think it's a clever solution with minimal changes. I want to look into how complicated it would be to do it "properly". Maybe we do it this way now & the complicated way later.

[jtraglia]

Fixed in #317. Closing.