Fix Windows in CI

[StanFromIreland]

https://vcpkg.io/en/package/libiconv.html

[StanFromIreland]

I've managed to install it, now I need to link it.

[StanFromIreland]

It now builds successfully, but does not import.

[bodograumann]

Great progress!

[StanFromIreland]

We can also try cache the dependency, so that we can skip the ~5 minute install step.

[StanFromIreland]

Ok, I have now got it to the same failures as on the MacOS runner!

[StanFromIreland]

Some progress has been made, it now builds on Windows and Mac alright, but it seems they don't provide all (and different) codecs:

 ======================================================================
ERROR: test_decode (test_iconvcodec.TestIconvcodecModule.test_decode)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/test_iconvcodec.py", line 17, in test_decode
    string = b"Hallo".decode("T.61")
             ^^^^^^^^^^^^^^^^^^^^^^^
LookupError: unknown encoding: T.61

======================================================================
ERROR: test_encode (test_iconvcodec.TestIconvcodecModule.test_encode)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/test_iconvcodec.py", line 8, in test_encode
    bytestring = "Hallo".encode("T.61")
                 ^^^^^^^^^^^^^^^^^^^^^^
LookupError: unknown encoding: T.61

======================================================================
ERROR: test_incremental_decode (test_iconvcodec.TestIconvcodecModule.test_incremental_decode)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/test_iconvcodec.py", line 34, in test_incremental_decode
    decoder = codecs.getincrementaldecoder("UCS2")()
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen codecs>", line 1013, in getincrementaldecoder
LookupError: unknown encoding: UCS2

======================================================================
ERROR: test_transliterate (test_iconvcodec.TestIconvcodecModule.test_transliterate)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/iconvcodec.py", line 13, in _iconv_encode_impl
    return encoder.iconv(msg, bufsize), len(msg)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
iconv.error: ('Illegal byte sequence', 92, 7, b'abc ss ')

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/test_iconvcodec.py", line 22, in test_transliterate
    bytestring = string.encode("ASCII//TRANSLIT")
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/runner/work/python-iconv/python-iconv/iconvcodec.py", line 96, in encode
    return _iconv_encode_impl(encoder, msg, errors)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/runner/work/python-iconv/python-iconv/iconvcodec.py", line 31, in _iconv_encode_impl
    raise UnicodeError(*e.args)
UnicodeError: ('Illegal byte sequence', 92, 7, b'abc ss ')
encoding with 'ASCII//TRANSLIT' codec failed

======================================================================
FAIL: test_encode_with_long_out (test_iconvcodec.TestIconvcodecModule.test_encode_with_long_out)
Edge case where output has more bytes than input as utf-8
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/runner/work/python-iconv/python-iconv/test_iconvcodec.py", line 14, in test_encode_with_long_out
    self.assertEqual(bytestring, b"(TM)")
AssertionError: b'TM' != b'(TM)'

[StanFromIreland]

@pffang, maybe you could please shine some light onto this, why does libiconv-for-Windows provide different codecs?

[StanFromIreland]

Patching tests, the CI is now fully green.

[pffang]

Sorry for late to see this.
I didn't get your points. But if you want to make tests, it's better to read the test strings from independent txt files.
Because the string encoding in code sources always depends on the encoding of source files and the processing methods of code language.
The compiler or interpreter may make some conversion, it's difficult to keep the encoding as your imagine.