19 tests fail

[yurivict]

See log here.

Version: 1.2.0
Python-3.9
FreeBSD 13.1

[newville]

@yurivict Hm, our tests all pass with Python 3.9 (and have for a long time). See these results.

Your log suggests that a great many (perhaps all - I may not have read the entire log closely) of the failures you see are associated with the error message:

  AttributeError: module 'dill' has no attribute 'dumps'

which means that a) there is an importable dill module, and b) it does not have a dumps function. Our code does assume that if dill imports, then calling dill.dumps() should work. The "basic usage" section of the dill documentation describes using dill.dumps() in its first use of dill, so I doubt we will change that assumption.

It seems more likely to me that there is something wrong with the installation of dill in your Python environment.
I don't know why that would be the case.

[yurivict]

Please note that pyproject.toml doesn't list dill in requires.

[yurivict]

The ```dill`` module was in fact broken. Updating it corrected all errors, except for this one:

=============================================================================================== FAILURES ================================================================================================
_________________________________________________________________________________________ test_Model_get_state __________________________________________________________________________________________

gmodel = <lmfit.Model: Model(gaussian)>

    def test_Model_get_state(gmodel):
        """Test for Model class function _get_state."""
        out = gmodel._get_state()
    
        assert isinstance(out, tuple)
        assert out[1] == out[2] is None
>       assert (out[0][1] is not None) == lmfit.jsonutils.HAS_DILL
E       AssertionError: assert (<function gaussian at 0x969ea8a60> is not None) == False
E        +  where False = <module 'lmfit.jsonutils' from '/usr/ports/math/py-lmfit/work-py39/lmfit-1.2.0/lmfit/jsonutils.py'>.HAS_DILL
E        +    where <module 'lmfit.jsonutils' from '/usr/ports/math/py-lmfit/work-py39/lmfit-1.2.0/lmfit/jsonutils.py'> = lmfit.jsonutils

tests/test_model.py:206: AssertionError

[newville]

@yurivict That error still seems related to dill. That test passes all our CI on all versions and platforms we test on.
Yes, dill is not listed in requirements.txt, because it is not required by lmfit. If it is installed, it might be used (and be expected to work correctly).

We really can not guarantee that our unit tests pass for everyone on every system and environment. That is not actually the purpose of our tests (nor is it really the purpose or promise of any test framework). We try to be helpful to anyone using our code, but we also are going to have a default view that tests that pass in our test environments but fail on some particular system or environment mean that environment is faulty.

Please use the results from running our tests on your environment as a test your environment.

[yurivict]

lmfit.jsonutils.HAS_DILL is False, while the latest version of dill (py39-dill-0.3.6_1) is installed.
This seems to be wrong.

[newville]

@yurivict Well, the code there (see

lmfit-py/lmfit/jsonutils.py

Line 9 in a03ac0f

try:

) reads:

try:
    import dill
    HAS_DILL = True
except ImportError:
    HAS_DILL = False

if HAS_DILL is False, then there was an ImportError with import dill.

Again, all of our tests pass on all the systems and all of the test environments. A failure on a different environment strongly suggests a problem with that environment.

This really is not an issue with lmfit. You ignored all of our requests and actual requirements for how to submit an issue and continue to suggest there is a problem here. We have been exceedingly polite even in the face of your inability to follow our request or reason about debugging your system.