feat(tests): Additional EIP-7934 test cases

[fselmo]

🗒️ Description

• Add test case for RLP block size at limit with all transaction types, including blobs, access lists, and authorizations
• Add test case for RLP block size at limit with logs emitted

🔗 Related Issues or PRs

Related to #1772

TODO:

• Refactor built transactions for with_all_tx_types into something that can be more useful across the board. I'm thinking default transactions that each fork can define if it includes a new tx type and tests can automatically use these transactions, override them as fixtures, or some other more useful flow than re-building every time.
  • This was done via with_all_typed_transactions indirect (new feature) marker

✅ Checklist

• All: Ran fast tox checks to avoid unnecessary CI fails, see also Code Standards and Enabling Pre-commit Checks:

  uvx --with=tox-uv tox -e lint,typecheck,spellcheck,markdownlint

• All: PR title adheres to the repo standard - it will be used as the squash commit message and should start type(scope):.
• All: Considered adding an entry to CHANGELOG.md.
• All: Considered updating the online docs in the ./docs/ directory.
• All: Set appropriate labels for the changes (only maintainers can apply labels).
• Tests: Ran mkdocs serve locally and verified the auto-generated docs for new tests in the Test Case Reference are correctly formatted.
• Tests: For PRs implementing a missed test case, update the post-mortem document to add an entry the list.

[fselmo]

Hmm having a difficult time reproducing this CI error locally. I added the concept of a @with_all_typed_transactions as a fixture that uses the tx_type covariant with fork to build default transactions. I successfully implemented this in the test that I wanted for testing the rlp block size limit. Each default transaction can be overridden if desired with:

@pytest.fixture
def type_2_default_transaction(sender):
    return Transaction(...)

Something in the stdout is being changed and ironically I am on a macos and I pass this test locally but the CI does not. I will keep looking into this but it would be nice to get a review here on the implementation and if anyone has an idea of what's causing this error, lmk.

[marioevz]

Awesome! Thanks for implementing this.

I wrote a couple of comments that might be worth considering.

[marioevz]

I think we could create ./tests/conftest.py and move all of these fixtures there with some minor changes to not require a change in the plugins:

Suggested change

	@pytest.fixture
	def typed_transaction(request, fork):
	@pytest.fixture
	def typed_transaction(tx_type: int, fork: Fork) -> Transaction:

Then the test adds typed_transaction: Transaction as a parameter, but it might have the downside that the test has to also include the tx_type: int parameter to simply not use it, and maybe some other downside I could be missing.

[fselmo]

Ok yeah, this was the route I was originally going to take but wasn't sure why a global conftest.py didn't exist yet and thought it may not be desirable. I think if you're good with that I am good with it as well 👍🏼

[fselmo]

it might have the downside that the test has to also include the tx_type: int parameter to simply not use it

I implemented it so that the typed_transaction fixture uses the tx_type and so all you have to do is use the @pytest.mark.with_all_tx_types and you can use the typed_transaction directly. If it's better to keep this as a separate marker, just let me know, but I think if we document it well enough that you can use either the tx_type or the typed_transaction then it's fairly frictionless and versatile.

lmk what you think about that or if the separation of concerns is better.

[fselmo]

Ok. So based on @danceratopz comments, I moved it back to the plugins. But based on your request @marioevz, I put this inside shared and available to shared/execute_fill.py plugin. I refactored the logic for the indirect covariant checks so it could be used by other markers in the future if needed. The CovariantDecorator takes an indirect kwarg that will run the values of the fork_attribute_name in this case through the fixtures (argnames) which is then what is used in the test.

I ended up creating a more global sender and I did so inside execute_fill.py but perhaps this can live somewhere else or this may not be ideal, so let me know there. But we can remove the redundant re-definitions of sender fixture that don't need some special type of logic if we have this more global one set here.

Would love thoughts on this from you both.

[fselmo]

Global sender allows us to remove quite a few redundancies in tests/ fixtures, as per the last commit in this PR.

[fselmo]

I'm just missing the documentation here which I can update but we should get consensus on the code first and then I can document the changes appropriately and add the release notes.