feat(testing/fill): Account-hash-based deploy addresses #2139

marioevz · 2026-02-04T22:40:51Z

🗒️ Description

Enhancement 1: Use account hashes to determine contract addresses

TL;DR

Reduce account duplication during pre-alloc grouping by making contract addresses a function of the account’s contents.

Details

Normally, account collisions are undesirable because once a test touches an account, its nonce increases, its storage may be modified, and its balance can change.

This is particularly relevant during test execution, where we deploy fresh contracts on a live network to start each test from the most deterministic state possible.

However, this constraint does not apply to tests that are filled and then consumed, since test execution always starts from a clean state.

In the case of Engine X tests, we always perform a forkchoice-update back to genesis. As a result, it does not matter if the same contract is used by multiple tests, because reverting to genesis rolls back all state changes. We can therefore safely reuse the same account across tests, provided it has the same initial state.

This PR modifies the behavior of Pre to calculate a contract’s address as a hash of the account’s contents, including its initial balance, nonce, storage, and code.

We also add a salt to the calculated hash to support cases where a single test requires multiple identical copies of the same account in the state.

Examples

Same contract, two different tests:

def test_1(pre: Alloc) -> None:
    contract_test_1 = pre.deploy_contract(code=Op.STOP)

def test_2(pre: Alloc) -> None:
    contract_test_2 = pre.deploy_contract(code=Op.STOP)

Outcomes

contract_test_1 == contract_test_2

Same contract, two different tests, multiple times:

def test_1(pre: Alloc) -> None:
    contract_test_1_a = pre.deploy_contract(code=Op.STOP)
    contract_test_1_b = pre.deploy_contract(code=Op.STOP)

def test_2(pre: Alloc) -> None:
    contract_test_2_a = pre.deploy_contract(code=Op.STOP)
    contract_test_2_a = pre.deploy_contract(code=Op.STOP)

Outcomes

contract_test_1_a == contract_test_2_a
contract_test_1_b == contract_test_2_b
contract_test_1_a != contract_test_1_b
contract_test_2_a != contract_test_2_b

Same contract, two different tests, different properties:

def test_1(pre: Alloc) -> None:
    contract_test_1 = pre.deploy_contract(code=Op.STOP, storage={0: 1})

def test_2(pre: Alloc) -> None:
    contract_test_2 = pre.deploy_contract(code=Op.STOP, storage={0: 2})

Outcomes

contract_test_1 != contract_test_2

EOA, no arguments, two different tests:

def test_1(pre: Alloc) -> None:
    sender_1 = pre.fund_eoa()

def test_2(pre: Alloc) -> None:
    sender_2 = pre.fund_eoa()

Outcomes

sender_1 == sender_2

EOA, same fund amount, two different tests:

def test_1(pre: Alloc) -> None:
    sender_1 = pre.fund_eoa(amount=0)

def test_2(pre: Alloc) -> None:
    sender_2 = pre.fund_eoa(amount=0)

Outcomes

sender_1 == sender_2

EOA, different fund amount, two different tests:

def test_1(pre: Alloc) -> None:
    sender_1 = pre.fund_eoa(amount=0)

def test_2(pre: Alloc) -> None:
    sender_2 = pre.fund_eoa(amount=1)

Outcomes

sender_1 != sender_2

Enhancement 2: Detect Grouping Separation Automatically

TL;DR

Automatically detect changes in the pre that would result in conflicts with other tests and put the test into a different group without having to rely on pytest.mark.pre_alloc_group.

Details

Keeps track of the addresses that were modified in the following ways:

pre[address] = Account(...)
pre.fund_address(...)
pre.deploy_contract(..., address=0x1234..., ...)
del pre[address]

All of these actions allow the test to directly modify an account at an specific address' contents, which would affect other tests.

We keep track of the addresses that were modified this way and hash the contents of each account to add it to the pre-alloc group hashing.

Reasoning behind pre.fund_address(...) in particular is that most tests expect other addresses (such as system addresses or precompiles, or even created contracts) to start with zero balance. This is different from specifying a non-zero balance in the pre.deploy_contract because this modifies the resulting address due to the hashing, which makes the collision impossible.

All usages of pytest.mark.pre_alloc_group have been removed from tests and pre-alloc grouping test execution of all Amsterdam test did not result in any issues. The marker itself was not removed just to retain the ability to group tests manually.

🔗 Related Issues or PRs

N/A.

✅ Checklist

All: Ran fast tox checks to avoid unnecessary CI fails, see also Code Standards and Enabling Pre-commit Checks:
```
uvx tox -e static
```
All: PR title adheres to the repo standard - it will be used as the squash commit message and should start type(scope):.
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.
Ported Tests: All converted JSON/YML tests from ethereum/tests or tests/static have been assigned @ported_from marker.

Cute Animal Picture

danceratopz · 2026-02-05T13:05:06Z

In the case of Engine X tests, we always perform a forkchoice-update back to genesis. As a result, it does not matter if the same contract is used by multiple tests, because reverting to genesis rolls back all state changes. We can therefore safely reuse the same account across tests, provided it has the same initial state.

This is not strictly necessary and is currently skipped in #1964 see https://github.com/danceratopz/execution-specs/blob/6fe382c0370ce94a935f65c89eac1073bc6324b6/packages/testing/src/execution_testing/cli/pytest_commands/plugins/consume/simulators/simulator_logic/test_via_engine.py#L43-L76.

We can certainly benchmark to see how much time is saved by avoiding the initial FCU!

But the initial plan was to skip it unless it was considered necessary for intermittent garbage collection (e.g. benchmarking efforts hinted that this would be necessary for Besu). A follow-up PR was to add --enginex-initial-fcu-freq to enable this; as-is it is entirely skipped.

Edit: We don't FCU post-genesis, but the effect is the same (as we don't FCU after the test either) - we're always building on top of genesis in engine x :)

spencer-tb

Smart optimization. Assuming this will save memory during phase 1 and intial stage of phase 2 filling. The salt mechanism is cool.

spencer-tb · 2026-02-05T14:04:41Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/filler/pre_alloc.py

        If the address is already present in the pre-alloc the amount will be
        added to its existing balance.


Docstring update suggestion

Suggested change

If the address is already present in the pre-alloc the amount will be

added to its existing balance.

Add a funded account to the pre-allocation.

The address must not already exist in the pre-allocation. To set the

balance of an account, use the `amount` parameter in `fund_eoa()` or

the `balance` parameter in `deploy_contract()` at creation time.

Added, thanks!

spencer-tb · 2026-02-05T14:06:54Z

packages/testing/src/execution_testing/base_types/composite_types.py

+class FrozenStorage(Storage):
+    """Frozen storage."""


Assuming will be used later.

Removed as it ended up not being necessary 👍

marioevz · 2026-02-09T01:51:14Z

@danceratopz @spencer-tb I added another very nice enhancement that virtually makes the pytest.mark.pre_alloc_group marker obsolete (🎉 ), please take a look and let me know what you think!

codecov · 2026-02-09T02:05:20Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.07%. Comparing base (342c7bc) to head (ce17f48).
⚠️ Report is 6 commits behind head on forks/amsterdam.

Additional details and impacted files

@@               Coverage Diff                @@
##           forks/amsterdam    #2139   +/-   ##
================================================
  Coverage            86.07%   86.07%           
================================================
  Files                  599      599           
  Lines                39472    39472           
  Branches              3780     3780           
================================================
  Hits                 33977    33977           
  Misses                4862     4862           
  Partials               633      633

Flag	Coverage Δ
unittests	`86.07% <ø> (ø)`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

marioevz · 2026-02-09T13:56:12Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/execute/pre_alloc.py

-    ) -> None:
-        """Set account associated with an address."""
-        raise ValueError(
-            "Tests are not allowed to set pre-alloc items in execute mode"


To reviewers: This check has been moved to SharedAlloc in the form of flags (see alloc_flags fixture for the skipping logic).

marioevz · 2026-02-09T15:23:09Z

tests/berlin/eip2930_access_list/test_acl.py

    )
    sender = pre.fund_eoa()
    tx_value = 1
-    pre.fund_address(sender, tx_value)


Note to reviewers: execute command handles this automatically since #1822 and for fill I think we should simply increment the default value if for some reason we come up with a test that requires more balance.

marioevz · 2026-02-09T15:28:15Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/filler/pre_alloc.py

+        if self._deleted_addresses:
+            buffer += b"\1"
+            for deleted_address in sorted(self._deleted_addresses):
+                buffer += deleted_address


Note to reviewers: We could even completely ban this behavior and make this change a bit cleaner. In cases where we "remove" system contracts from the state we don't really remove them, we instead do pre[SYSTEM_CONTRACT_ADDRESS] = Account(<empty account>) to overwrite it.

fselmo

This is a lot to review but I think it looks good. Really clever change that makes a lot of sense 🔥.

It's really difficult to know though, imo, if this will create any friction with execute remote. I think this is "less crucial" of a way to test (and more rnd) and we can certainly fix any quirks that bubble up after this PR is in... but I do wonder if there aren't some basic sanity check unit tests we can write for some of these methods that guarantee we are not introducing bugs to this functionality. I think this would be pretty difficult though... curious if others have any ideas here.

I asked Claude to look at this and I think there's a potential bug with setting self[address] instead of using __internal_set_item__(address, ...). I couldn't reproduce it in kurtosis bc I think it'd only be present in some very unique scenario... but asking here to make sure.

Everything else looks pretty good to me. This will be very nice to get in 👍🏼

fselmo · 2026-02-09T21:28:23Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/execute/pre_alloc.py

            account = self[address]
            if account is not None:
-                account.balance = ZeroPaddedHexNumber(new_balance)
+                self[address] = account.copy(balance=new_balance)


Clarifying q... doesn't this also need to use __internal_set_item__(address, ...) in order to actually run through the RPC funding logic?

It doesn't seem easy, but do you think it's possible to write basic, sanity checking unit tests to make sure these execute mode expectations don't break? It would make reviews easier to have confidence in imo as it's not so straightforward / quick to test execute remote things as a reviewer.

fselmo · 2026-02-09T22:12:22Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/execute/pre_alloc.py

                    account = self[address]
                    if account is not None:
-                        account.balance = ZeroPaddedHexNumber(current_balance)
+                        self[address] = account.copy(balance=current_balance)


Same thought here about using __internal_setitem__()

tests/prague/eip7702_set_code_tx/test_set_code_txs.py

fselmo · 2026-02-09T22:29:25Z

packages/testing/src/execution_testing/cli/pytest_commands/plugins/shared/pre_alloc.py

+        """
+        Deploy a contract to the allocation.
+
+        Warning: `address` parameter is a temporary solution to allow tests to


Stray thought... I actually only now realized this is supposed to be temporary. It looks like only one test uses this and it makes sense. Would this be a good time to change this to something like _address (private attr... do not use unless you know what you're doing kind of thing?). I feel like it may be easy for a test that uses this to sneak by if it's a public property on pre.deploy_contract().

fselmo · 2026-02-09T22:37:43Z

packages/testing/src/execution_testing/base_types/composite_types.py


+    model_config = {
+        **CamelModel.model_config,
+        "frozen": True,


Nothing to add here. Just a thought that since this is frozen I think we could probably cache the hash in case we ever end up calling it more than once we don't end up hashing it multiple times. I'm pretty sure the happy path in this PR only calls it once though so consider this just a nit.

Co-authored-by: felipe <fselmo2@gmail.com>

spencer-tb reviewed Feb 5, 2026

View reviewed changes

marioevz added 18 commits February 9, 2026 01:10

feat(testing/fill): Account-hash-based deploy addresses

a3d1832

fix(testing): Allow EOAs to have code for collisions

90582d8

fix(tests): All tests using pre.fund_address

0c54a2e

fix(tests): Tests setting EOA code

cf0c138

fixes

6916724

Fixes

5c7dabc

Add correct markers to all tests

c1bab60

fix(testing): test generator use appropriate mark

d4db27e

fix(testing/filler): Logic

998b359

refactor(tests): Remove most pre-alloc-group markers

f8931a3

feat(filler): Make pre-alloc auto-groupable

ed7cac3

fix: bug, add unit test

f097119

refactor: Remove pre_fund_address marker

4a20333

refactor: Remove extra flags

68e96bc

fix: execute

c208721

refactor: Move helper methods

1bed237

fix: Separators

1ce18f9

fix: Static tests

2c7576c

marioevz force-pushed the minimize-pre-allocations-on-grouping branch from b9697da to 2c7576c Compare February 9, 2026 01:17

fix: Tox, review comments

936d7bd

marioevz requested a review from spencer-tb February 9, 2026 01:49

marioevz marked this pull request as ready for review February 9, 2026 01:49

fix(tests): Mark test_creates_collisions

506ca36

marioevz commented Feb 9, 2026

View reviewed changes

fselmo reviewed Feb 9, 2026

View reviewed changes

Update tests/prague/eip7702_set_code_tx/test_set_code_txs.py

ce17f48

Co-authored-by: felipe <fselmo2@gmail.com>

		If the address is already present in the pre-alloc the amount will be
		added to its existing balance.

-        If the address is already present in the pre-alloc the amount will be
-        added to its existing balance.
+        Add a funded account to the pre-allocation.
+        The address must not already exist in the pre-allocation. To set the
+        balance of an account, use the `amount` parameter in `fund_eoa()` or
+        the `balance` parameter in `deploy_contract()` at creation time.

feat(testing/fill): Account-hash-based deploy addresses #2139

Are you sure you want to change the base?

feat(testing/fill): Account-hash-based deploy addresses #2139

Uh oh!

Conversation

marioevz commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🗒️ Description

Enhancement 1: Use account hashes to determine contract addresses

TL;DR

Details

Examples

Same contract, two different tests:

Outcomes

Same contract, two different tests, multiple times:

Outcomes

Same contract, two different tests, different properties:

Outcomes

EOA, no arguments, two different tests:

Outcomes

EOA, same fund amount, two different tests:

Outcomes

EOA, different fund amount, two different tests:

Outcomes

Enhancement 2: Detect Grouping Separation Automatically

TL;DR

Details

🔗 Related Issues or PRs

✅ Checklist

Cute Animal Picture

Uh oh!

danceratopz commented Feb 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

spencer-tb left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

marioevz commented Feb 9, 2026

Uh oh!

codecov bot commented Feb 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fselmo left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fselmo Feb 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

marioevz commented Feb 4, 2026 •

edited

Loading

danceratopz commented Feb 5, 2026 •

edited

Loading

codecov bot commented Feb 9, 2026 •

edited

Loading

fselmo Feb 9, 2026 •

edited

Loading