Failure in test-python-utils with v2.6.0a2

[santoshshanbhogue]

Problem description
Two tests fail on my mac (there are others that fail, but they are comments in the source code that these are expected such as in test_mixture.py::TestMixture::test_equilibrate2)

Steps to reproduce

run scons test test-python-utils

Behavior

See the screen dump at the end of this message for more info

System information

• Cantera version: 2.6.0a2 commit #627e21e95
• OS: Mac OS 10.15.7
• Python/MATLAB/other software versions: 3.7.4

Attachments

../../build/python/cantera/test/test_transport.py .......................................                        [ 96%]
../../build/python/cantera/test/test_utils.py ..................FF.......                                        [100%]

======================================================= FAILURES =======================================================
______________________________________________ TestPyToAnyValue.test_none ______________________________________________

self = <cantera.test.test_utils.TestPyToAnyValue testMethod=test_none>

    def test_none(self):
>       self.check_conversion(None, "void")

../../build/python/cantera/test/test_utils.py:30: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../../build/python/cantera/test/test_utils.py:14: in check_conversion
    self.assertEqual(held_type, check_type)
E   AssertionError: 'v' != 'void'
E   - v
E   + void
__________________________________________ TestPyToAnyValue.test_scalar_bool ___________________________________________

self = <cantera.test.test_utils.TestPyToAnyValue testMethod=test_scalar_bool>

    def test_scalar_bool(self):
>       self.check_conversion(True, "bool")

../../build/python/cantera/test/test_utils.py:55: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../../build/python/cantera/test/test_utils.py:14: in check_conversion
    self.assertEqual(held_type, check_type)
E   AssertionError: 'b' != 'bool'
E   - b
E   + bool

Additional context

[ischoegl]

@santoshshanbhogue ... thank you for reporting this issue. Could you please provide some additional information about your specific setup? (both hardware and cantera configuration). I am asking as there are some apparent differences between your system and the GitHub OSX test environment (where those tests pass).

[santoshshanbhogue]

Hardware info:
Model Name: MacBook Pro
Model Identifier: MacBookPro15,1
Processor Name: 8-Core Intel Core i9
Processor Speed: 2.3 GHz
Number of Processors: 1
Total Number of Cores: 8
L2 Cache (per Core): 256 KB
L3 Cache: 16 MB
Hyper-Threading Technology: Enabled
Memory: 32 GB

Cantera configuration:
prefix = '/Users/santosh1/packages/ct_dev_version'
python_package = 'full'
matlab_toolbox = 'y'
matlab_path = '/Applications/MATLAB_R2019a.app/'
f90_interface = 'n'
system_eigen = 'n'
system_fmt = 'n'
system_sundials = 'n'

System kernel info:
Darwin Kernel Version 19.6.0

Let me know if you need anything else

[bryanwweber]

I think the NumPy, Cython, and Python versions would all help too.

[santoshshanbhogue]

I'm using Python 3.7.4
Numpy '1.17.2'
and cython '0.29.13'

[ischoegl]

Interesting - I was mostly curious about Apple Silicon vs Intel processor. I don’t use a Mac and there’s nothing out of the ordinary, so I don’t have any ideas at the moment.

[speth]

I'm mostly interested in which C++ compiler C++ standard library version are being used, and what version of Boost is present. One of those is responsible for the strings that are being returned here.

I think what's happening is that we've ended up setting some very specific expectations for the return values of AnyValue::type_str() in these tests, when it's really just meant to provide something to put in an error message.

[santoshshanbhogue]

I'm using g++ and compiling with --std=c++0x

Apple clang version 12.0.0 (clang-1200.0.32.29)
The standard library version is 201103
Boost version: 1.55.0

[speth]

It's interesting that this is (I think) the version of Boost that first introduced the demangle function that we're using here. Can you check the top of your scons build output for the lines that mention Boost and the check for the dmangle function. For example, I have:

INFO: Found Boost version 1.71
Checking whether boost::core::demangle is declared... (cached) yes

p.s. You aren't actually using g++. As you can see from the version output mentioning clang, this is just a very confusing alias that Apple has introduced.

[santoshshanbhogue]

Interesting, actually, no, the 1.55.0 version that I use does not have the demangle function and the scons build log indicates as such

I just upgraded to v1.76 (the latest) which has the demangle function and all tests pass

I also just checked (https://cantera.org/compiling/dependencies.html#other-required-software) and it states that Boost > 1.54 will suffice

By the way, is there some plan for development in MATLAB for which you need contributors?

[speth]

We aren't deliberately requiring a newer Boost version, and apparently the note I made in SConstruct that demangle is available in 1.55 isn't correct.

I'd like to keep this open, as there are apparently a couple of little hiccups here that warrant fixes.

[bryanwweber]

I looked into which version has core/demangle.hpp and it seems to be in 1.56.0.