Difficulty running bindc_03.py in regular CPython

[rebcabin]

I thought I had the PYTHONPATH figured out, but CType can't be found.

PYTHONPATH=/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/integration_tests/../src/runtime/lpython:/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/integration_tests/.. python bindc_03.py
Traceback (most recent call last):
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/integration_tests/bindc_03.py", line 12, in <module>
    def g(a: CPtr, value: i32, offset_value: bool) -> None:
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/src/runtime/lpython/lpython.py", line 477, in wrap
    func = CTypes(func, c_shared_lib, c_shared_lib_path)
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/src/runtime/lpython/lpython.py", line 369, in __init__
    self.cf = CTypes.emulations[self.name]
AttributeError: type object 'CTypes' has no attribute 'emulations'

[certik]

The above way works for most cases, but does not work if you call into C, as you need to tell the emulation layer where to find it.

$ PYTHONPATH=../src/runtime/lpython python bindc_03.py 
Traceback (most recent call last):
  File "/Users/ondrej/repos/lpython/integration_tests/bindc_03.py", line 12, in <module>
    def g(a: CPtr, value: i32, offset_value: bool) -> None:
  File "/Users/ondrej/repos/lpython/src/runtime/lpython/lpython.py", line 477, in wrap
    func = CTypes(func, c_shared_lib, c_shared_lib_path)
  File "/Users/ondrej/repos/lpython/src/runtime/lpython/lpython.py", line 369, in __init__
    self.cf = CTypes.emulations[self.name]
AttributeError: type object 'CTypes' has no attribute 'emulations'

For that you also have to set LPYTHON_PY_MOD_NAME and LPYTHON_PY_MOD_PATH. It's not user friendly currently, but it does work. You need consult CMakeLists.txt for the details how it does it.

[rebcabin]

got it

[rebcabin]

└─(12:33:29 on vector-backend)──> PYTHONPATH=".:../src/runtime/lpython:.." LPYTHON_PY_MOD_NAME="" LPYTHON_PY_MOD_PATH="_lpython-tmp-test-cpython" python bindc_03.py
Traceback (most recent call last):
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/integration_tests/bindc_03.py", line 12, in <module>
    def g(a: CPtr, value: i32, offset_value: bool) -> None:
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/src/runtime/lpython/lpython.py", line 477, in wrap
    func = CTypes(func, c_shared_lib, c_shared_lib_path)
  File "/Users/brian/Dropbox/Mac/Documents/GitHub/lpython/src/runtime/lpython/lpython.py", line 367, in __init__
    self.cf = self.library[self.name]
  File "/Users/brian/miniforge3/envs/lp/lib/python3.10/ctypes/__init__.py", line 392, in __getitem__
    func = self._FuncPtr((name_or_ordinal, self))
AttributeError: dlsym(0x87869030, g): symbol not found
(lp) ┌─(...uments/GitHub/lpython/integration_tests)───(brian@MacBook-Pro:s001)─┐
└─(12:35:23 on vector-backend)──>

[rebcabin]

I'm leaving this open. bindc_03.py doesn't work in CPython, but most of the other bindc programs do work with the ENVIRONMENT variables set as above. I verified by hand

    Start 22: bindc_01
1/8 Test #22: bindc_01 .........................   Passed    0.03 sec
    Start 23: bindc_02
2/8 Test #23: bindc_02 .........................   Passed    0.06 sec
    Start 24: bindc_07
3/8 Test #24: bindc_07 .........................   Passed    0.07 sec
    Start 25: bindc_08
4/8 Test #25: bindc_08 .........................   Passed    0.06 sec
    Start 26: bindc_09
5/8 Test #26: bindc_09 .........................   Passed    0.06 sec
    Start 27: bindc_09b
6/8 Test #27: bindc_09b ........................   Passed    0.06 sec
    Start 28: bindc_10
7/8 Test #28: bindc_10 .........................   Passed    0.06 sec
    Start 29: bindc_11
8/8 Test #29: bindc_11 .........................   Passed    0.06 sec

[certik]

Yes, it's a bug: it works in LPython, but fails in CPython. So either we need to disable it in LPython with an error message, or get it working with CPython. It's low priority and for now let's keep the LPython test, since at least LPython will keep working for this, and we can get the CPython version working later. We can always later decide to just give an error message in LPython for this case if we do not want to support it.