bug: deltakit-decode tests fails on DeprecationWarning

[nelimee]

Bug Description

See this job failing:

FAILED deltakit-decode/tests/unit/utils/test_derivation_tools.py::TestDerivationTools::test_generate_expectation_data_throws_warn_if_both_bool_true[samples0] - DeprecationWarning: This process (pid=2338) is multi-threaded, use of fork() may lead to deadlocks in the child.
FAILED deltakit-decode/tests/unit/utils/test_derivation_tools.py::TestDerivationTools::test_generate_expectation_values_raises_warning_if_num_processes_0_or_less_than_negative_1[0] - DeprecationWarning: This process (pid=2338) is multi-threaded, use of fork() may lead to deadlocks in the child.
FAILED deltakit-decode/tests/unit/utils/test_derivation_tools.py::TestDerivationTools::test_generate_expectation_values_raises_warning_if_num_processes_0_or_less_than_negative_1[-2] - DeprecationWarning: This process (pid=2338) is multi-threaded, use of fork() may lead to deadlocks in the child.
FAILED deltakit-decode/tests/unit/utils/test_derivation_tools.py::TestDerivationTools::test_generate_expectation_values_raises_warning_if_num_processes_0_or_less_than_negative_1[-3] - DeprecationWarning: This process (pid=2338) is multi-threaded, use of fork() may lead to deadlocks in the child.
FAILED deltakit-decode/tests/unit/utils/test_derivation_tools.py::TestDerivationTools::test_generate_expectation_values_raises_warning_if_num_processes_0_or_less_than_negative_1[-10] - DeprecationWarning: This process (pid=2338) is multi-threaded, use of fork() may lead to deadlocks in the child.

Minimal Reproducible Example (MRE)

Make sure that all the tests are running when running pytest in the main deltakit repository by replacing

# deltakit/pyproject.toml
testpaths = ["./tests/"]

by

# deltakit/pyproject.toml
testpaths = [
  "./tests/",
  "./deltakit-circuit/tests/",
  "./deltakit-core/tests/",
  "./deltakit-decode/tests/",
  "./deltakit-explorer/tests/",
]

uv sync --all-packages --python 3.13 --resolution lowest-direct --group test
uv run --python 3.13 --no-sync pytest

Actual behaviour

With the above changes,

uv run --python 3.13 --no-sync pytest

fails but

uv run --python 3.13 --no-sync pytest deltakit-decode/tests

runs without any issue. This hints to tests in other deltakit sub-packages being part of the problem.

Expected behaviour

Tests are passing without any issue.

Additional context

No response

[motazmuhammad]

@nelimee if you do not mind I would like this one assigned to me.

[motazmuhammad]

Hi @nelimee First of all, thank you for accepting to assign this to me.

First, I have tried to reproduce the issue locally. It does not reproduce on windows, so I tried with WSL and it does reproduce. This seems to make sense as fork() is unix only.

Then I was able to pinpoint the problem to the file _derivation_tools.py which uses pathos for multiprocessing

To solve this I think it may make sense to explicitly set the multiprocessing start method to spawn which is generally safer. Therefore, adding the following line o the aforementioned file after import pathos pathos.helpers.mp.set_start_method("spawn", force=True)

seems to be the logical choice. I verified that locally, and it seems to eliminate the DeprecationWarning
if that aligns with what you had in mind. I will go ahead make the changes and initiate PR

[nelimee]

Hi!

I verified that locally, and it seems to eliminate the DeprecationWarning if that aligns with what you had in mind. I will go ahead make the changes and initiate PR

For this issue, I had nothing in mind really. I was expecting this to be a little bit more involved than that, so I am very happy to see that a small change should do it!

I am a little worried about the warnings around the multiprocessing function https://docs.python.org/3/library/multiprocessing.html#multiprocessing.set_start_method:

Set the method which should be used to start child processes. The method argument can be 'fork', 'spawn' or 'forkserver'. Raises RuntimeError if the start method has already been set and force is not True. If method is None and force is True then the start method is set to None. If method is None and force is False then the context is set to the default context.

Note that this should be called at most once, and it should be protected inside the if name == 'main' clause of the main module.

In particular the Note that this should be called at most once.

[motazmuhammad]

@nelimee I just wanted to mention that the point you raised is a good catch thank you for catching that. Now, I am looking at other alternatives and evaluating them. I have not fully finished the exploration stage yet. That is why I have not wrote any comments, but I will very soon.

[motazmuhammad]

@nelimee thank you for pointing the problem with set_start_method, after some investigation I realized that we could used a local context in _derivation_tools.py and that would fix the problem locally without any global unwanted side effects. The proposed solution now becomes
local_context = pathos.helpers.mp.get_context("spawn")
with local_context.Pool(num_processes) as p:

This would create a per call context with no global side effects. I tried that locally and it works as well. If that approach is okay. I will commit it.

[nelimee]

I think that's exactly what is needed to solve that issue! I was not aware of that context feature of multiprocessing, very nice catch! Pretty sure this is the solution, you can go ahead and submit a PR with this.

[motazmuhammad]

@nelimee i am sorry my personal computer is not operational so until i get another or have it fixed i cant contribute can you unassign me until this is fixed