feat(tests): Identity precompile test conversions

[reedsa]

🗒️ Description

Identity Precompile test conversions.

🔗 Related Issues

#972 (stPreCompiledContracts / stPreCompiledContracts2)

🗑️ ethereum/tests removals

• Remove all identity precompile tests tests#1482

✅ Checklist

• All: Set appropriate labels for the changes.
• All: Considered squashing commits to improve commit history.
• All: Added an entry to CHANGELOG.md.
• All: Considered updating the online docs in the ./docs/ directory.
• Tests: All converted JSON/YML tests from ethereum/tests have been added to converted-ethereum-tests.txt.
• Tests: A PR with removal of converted JSON/YML tests from ethereum/tests have been opened.
• Tests: Included the type and version of evm t8n tool used to locally execute test cases: e.g., ref with commit hash or geth 1.13.1-stable-3f40e65.
• Tests: Ran mkdocs serve locally and verified the auto-generated docs for new tests in the Test Case Reference are correctly formatted.

[reedsa]

@marioevz @winsvega There are a couple test cases failing, seeing differences in the expectations from the original tests. I have a couple questions about these cases, as follows:

• The original test for identity_1_nonzero, expects the call to succeed and for the stored value to be 0. However this implementation results in the call failing for some reason.
• The original test for identity_5 does not include the call result in the expected output. In this case, it looks like the call should revert but expects the stored value to be set as 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF. When running the converted test case I do not see the value being stored. Should the expected result just be 0xDEADBEEF based on this implementation, or is there a better way to capture the original test? Does this happen because the call reverts and the contract code does not complete?

[winsvega]

Have you checked the vmtrace of original tests?

You can run retesteth test.json --vmtrace

You can select vector --singlenet Cancun -d index

[reedsa]

Have you checked the vmtrace of original tests?

You can run retesteth test.json --vmtrace

You can select vector --singlenet Cancun -d index

Yeah the trace for the CALL in identity_5 shows "error" out of gas, but the specified memory is still allocated with that value. I'm wondering if since the call reverts from the error, if these new tests do not end up allocating that memory.

[winsvega]

If a subcall fails with OOG, the upper level call continues, thus stroing all the values, but the failed call returns 0.

Sorry I have not time to review, but I can give hints.

[winsvega]

Ah look

https://github.com/ethereum/tests/blob/develop/src/GeneralStateTestsFiller/stPreCompiledContracts2/CallIdentity_1_nonzeroValueFiller.json

In original test the pre storage is empty (zero)

Thus if the call fails we don't see it as the storage was not set to deadbeef
Thats why it is good to always set the storage to nonzero value if we expect to verify return code of 0.

[winsvega]

About identity5, if you see the original test makes little sense, it can be errors.

See if the test logic make sense, try to enhance it with more scenarios using pytest.

If you see that original test does not do anything useful, see what was the intent, maybe it can be fixed.

[winsvega]

Fill tests shows much more mismatch in the expectations. Need to check the logic, something is off.

[reedsa]

I realized I needed a different gas_limit for a couple of cases. Those now reside in a separate test.

[reedsa]

I was expecting the identity precompile tests to be valid from Frontier but I see failures when I try to fill these. Is there something extra needed to cover these?

The coverage failures look exactly as they did for the blake2 precompile tests. Adding that description to the thread below. #1244 (comment)

[reedsa]

Manual Coverage Analysis

/evmone/evmc/include/evmc - helpers.h ✅

     302 LBC           0 :     case EVMC_PRAGUE:
     303               0 :         return "Prague";

Lines lost due to coverage script not filling Prague for python tests, and since latest release of ethereum/tests contains prague, this is expected.

/evmone/test/state - mpt.cpp ✅

      95 LBC           0 :     static MPTNode ext(const Path& path, std::unique_ptr<MPTNode> child) noexcept
      96                 :     {
      97               0 :         assert(child->m_kind == Kind::branch);
      98               0 :         MPTNode node{Kind::ext, path};
      99               0 :         node.m_children[0] = std::move(child);
     100               0 :         return node;

...

     244 LBC           0 :     case Kind::ext:
     245                 :     {
     246               0 :         encoded = rlp::encode(m_path.encode(m_kind)) + encode_child(*m_children[0]);
     247               0 :         break;

Lost due how test accounts (in pre-alloc) are structured in EEST vs ethereum/tests, which is irrelevant.

[winsvega]

on Frontier the gas requirements could be different. the test in ethereum/tests is for Cancun only. see the version of it in ethereum/legacytests if it has a vertion for Frontier.

[winsvega]

@marioevz shall we add it to default codebase in Opcodes this struct could be useful

[reedsa]

@marioevz if you have a few mintues for a final look, I think this is ready to merge. Thanks!

[marioevz]

Awesome! Thanks for implementing these :)

I think it only needs a rebase for us to be able to delete the fillers from ./tests/static/state_tests/....

[reedsa]

@marioevz @winsvega I see the coverage is failing because the tests were removed (in this PR). Are those meant to kept around?

[winsvega]

No its a bug. Iam working on a fix