fix: make blake2b optional

[fubuloubu]

What was wrong?

fixes: #1959
may fix: #1941

How was it fixed?

Used the native implementation of blake2b compression function if blake2b-py is not installed.
Made blake2b-py an optional dependency via eth-extra.

To-Do

• Clean up commit history

• Add entry to the release notes

Cute Animal Picture

[cburgdorf]

LGTM

[cburgdorf]

Generally looks good but CI errors need to be addressed. Also just pinging @carver since that might affect Py-EVM and he's closer to that project than I am.

[fubuloubu]

Generally looks good but CI errors need to be addressed.

As far as I can tell, none of the linting errors have to do with files I edited.

Also just pinging @carver since that might affect Py-EVM and he's closer to that project than I am.

Yes, this definitely might affect downstream users of py-evm, like eth-tester and trinity. trinity has been deprecated, but I think for eth-tester it should be an optional dependency (and web3py[tester] as well)

[cburgdorf]

Sorry for the late response.

As far as I can tell, none of the linting errors have to do with files I edited.

Yes, but they still need to be addressed since we can not merge a PR that didn't pass all CI checks.

I'm not working on this project anymore (and I just actually noticed this is a Py-EVM PR not a blake2b PR 🙈 ) but I know that @carver is currently working on EIP-1559 support for Py-EVM so if you don't feel like investing time into fixing these linting issues you may just wait for them to be fixed in master and then rebase at a later point :)

[fubuloubu]

I just actually noticed this is a Py-EVM PR not a blake2b PR

Yes, this seemed like the best solution to this issue as adding Windows support to the Rust-based blake2-py would be much more difficult, and the blake2 precompile is not used nearly enough in production to justify that work.

@carver is currently working on EIP-1559 support for Py-EVM so if you don't feel like investing time into fixing these linting issues you may just wait for them to be fixed in master and then rebase

I would prefer this sooner than later, I will take a look at the linting issues. This is a blocker for me.

[cburgdorf]

Thanks @fubuloubu

[MatthiasLohr]

Great! Thanks!

Any idea, when this gets released (uploaded to PyPI)?

[carver]

The next release will probably happen after London tests pass. You can watch #2021

[fubuloubu]

Note: this will not fix #1941 because the issue there is the pyethash dependency