[2.3.x] CI: skip lxml encode test on Windows

[jorisvandenbossche]

The Windows CI has been failing for a couple of days on the 2.3.x branch, with one failing html test test_encode when using the lxml backend:

 ================================== FAILURES ===================================
_______________ TestReadHtml.test_encode[letz_latin1.html-lxml] _______________
[gw0] win32 -- Python 3.12.7 C:\Users\runneradmin\micromamba\envs\test\python.exe

self = <pandas.tests.io.test_html.TestReadHtml object at 0x0000022907CCDC30>
html_encoding_file = 'D:\\a\\pandas\\pandas\\pandas\\tests\\io\\data\\html_encoding\\letz_latin1.html'
flavor_read_html = functools.partial(<function read_html at 0x00000228FA7CD6D0>, flavor='lxml')

    @pytest.mark.filterwarnings(
        "ignore:You provided Unicode markup but also provided a value for "
        "from_encoding.*:UserWarning"
    )
    def test_encode(self, html_encoding_file, flavor_read_html):
        base_path = os.path.basename(html_encoding_file)
        root = os.path.splitext(base_path)[0]
        _, encoding = root.split("_")
    
        try:
            with open(html_encoding_file, "rb") as fobj:
>               from_string = flavor_read_html(
                    fobj.read(), encoding=encoding, index_col=0
                ).pop()

pandas\tests\io\test_html.py:1395: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
pandas\io\html.py:1240: in read_html
    return _parse(
pandas\io\html.py:983: in _parse
    tables = p.parse_tables()
pandas\io\html.py:249: in parse_tables
    tables = self._parse_tables(self._build_doc(), self.match, self.attrs)
pandas\io\html.py:791: in _build_doc
    r = parse(self.io, parser=parser)
C:\Users\runneradmin\micromamba\envs\test\Lib\site-packages\lxml\html\__init__.py:914: in parse
    return etree.parse(filename_or_url, parser, base_url=base_url, **kw)
src\\lxml\\etree.pyx:3589: in lxml.etree.parse
    ???
src\\lxml\\parser.pxi:1958: in lxml.etree._parseDocument
    ???
src\\lxml\\parser.pxi:1984: in lxml.etree._parseDocumentFromURL
    ???
src\\lxml\\parser.pxi:1887: in lxml.etree._parseDocFromFile
    ???
src\\lxml\\parser.pxi:1200: in lxml.etree._BaseParser._parseDocFromFile
    ???
src\\lxml\\parser.pxi:633: in lxml.etree._ParserContext._handleParseResultDoc
    ???
src\\lxml\\parser.pxi:743: in lxml.etree._handleParseResult
    ???
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

>   ???
E     File "<string>", line 0
E   lxml.etree.XMLSyntaxError: unknown error

src\\lxml\\parser.pxi:672: XMLSyntaxError

The test is already allowed to fail on Windows when the file uses UTF-16 or UTF-32 encoding, but so now it fails with latin1 encoding as well.

Checking a working vs failing build on 2.3.x, I do notice that libxml2 was updated from 2.12.7 to 2.13.4 when it started failing. But on the main branch, it is also using this latest version and there it is not failing ..

So not directly an idea what is going on, and cannot reproduce this locally. Given it is working on main, and for all other platforms (and we already skip certain encodings as well), I am inclined to just skip the test for Windows on 2.3.x.

cc @mroeschke in case you have seen something similar on main (maybe something we forgot to backport?)

[mroeschke]

Thanks. I haven't seen this failure on main; only on 2.3.x as you mentioned.