lxml test fails with libxml2 2.12

This is seen in Fedora too: https://bugzilla.redhat.com/show_bug.cgi?id=2251911

There is a typo in the document used in the test; the opening <Document> tag is missing the final right angle bracket, making the XML invalid. If I'm right, that typo is causing the problem. The method of handling invalid documents is something that can legitimately change between one version of a dependency and another, and it's not what I'm trying to test with that test case.

Please try the attached diff (or the head of the 'master' branch) and see if that solves the problem.

Thanks. Can confirm that fixes the issue.

I've tested that patch and with libxml 2.12.3 I have pytaest failing unit

+ PYTHONPATH=/home/tkloczko/rpmbuild/BUILDROOT/python-beautifulsoup4-4.12.2-6.fc35.x86_64/usr/lib64/python3.8/site-packages:/home/tkloczko/rpmbuild/BUILDROOT/python-beautifulsoup4-4.12.2-6.
+ /usr/bin/pytest -ra -m 'not network'
============================= test session starts ==============================
platform linux -- Python 3.8.18, pytest-7.4.3, pluggy-1.3.0
rootdir: /home/tkloczko/rpmbuild/BUILD/beautifulsoup4-4.12.2
collected 656 items

bs4/tests/test_builder.py ..... [ 0%]
bs4/tests/test_builder_registry.py ........... [ 2%]
bs4/tests/test_css.py .................................................. [ 10%]
.......... [ 11%]
bs4/tests/test_dammit.py .................................. [ 16%]
bs4/tests/test_element.py ..... [ 17%]
bs4/tests/test_formatter.py .............. [ 19%]
bs4/tests/test_fuzz.py .....ssssss [ 21%]
bs4/tests/test_html5lib.py ............................................. [ 28%]
...................................... [ 33%]
bs4/tests/test_htmlparser.py ........................................... [ 40%]
...........F.................... [ 45%]
bs4/tests/test_lxml.py ................................................. [ 52%]
.................................................... [ 60%]
bs4/tests/test_navigablestring.py ........ [ 62%]
bs4/tests/test_pageelement.py .................................... [ 67%]
bs4/tests/test_soup.py ................................................. [ 75%]
......... [ 76%]
bs4/tests/test_tag.py ....................... [ 79%]
bs4/tests/test_tree.py ................................................. [ 87%]
........................................................................ [ 98%]
........... [100%]

=================================== FAILURES ===================================
_____ TestHTMLParserTreeBuilder.test_smart_quotes_converted_on_the_way_in ______

self = <bs4.tests.test_htmlparser.TestHTMLParserTreeBuilder object at 0x7f1aa6965c40>

    def test_smart_quotes_converted_on_the_way_in(self):
        # Microsoft smart quotes are converted to Unicode characters during
        # parsing.
        quote = b"<p>\x91Foo\x92</p>"
        soup = self.soup(quote)
> assert soup.p.string == "\N{LEFT SINGLE QUOTATION MARK}Foo\N{RIGHT SINGLE QUOTATION MARK}"
E AttributeError: 'NoneType' object has no attribute 'string'

bs4/tests/__init__.py:808: AttributeError
=========================== short test summary info ============================
SKIPPED [6] bs4/tests/test_fuzz.py:60: html5lib problems
FAILED bs4/tests/test_htmlparser.py::TestHTMLParserTreeBuilder::test_smart_quotes_converted_...

@kloczek I am unable to reproduce that failure on NixOS with Python 3.11.6, libxml2 2.12.3 and libxslt 1.1.39.

It's very unlikely that the error you found in test_htmlparser.TestHTMLParserTreeBuilder is related to this patch, for two reasons:

1. The test that fails is testing Python's built-in HTML parser, not lxml. No lxml code should be running.
2. The patch only changes a test (in XMLTreeBuilderSmokeTest), not any code that would be run by the failing test.

In addition, I'm unable to reproduce the issue in my own Python 3.8 environment.

I'd be interested to know:

* Whether the TestHTMLParserTreeBuilder failure is reliably repeatable in your Python 3.8 environment, or whether it's intermittent.
* Whether backing out the patch fixes the TestHTMLParserTreeBuilder failure (at the expense of potentially introducing an lxml test failure)
* Whether installing lxml (and thus libxml2) into an environment that doesn't have the TestHTMLParserTreeBuilder failure, introduces the failure.