These are tests from lxml's ET compatibility test suite that now fail in ElementTree:

    def test_delslice_step(self):
        Element = self.etree.Element
        SubElement = self.etree.SubElement

        a = Element('a')
        b = SubElement(a, 'b')
        c = SubElement(a, 'c')
        d = SubElement(a, 'd')
        e = SubElement(a, 'e')

        del a[1::2]
        self.assertEquals(
            [b, d],
            list(a))

    def test_delslice_step_negative(self):
        Element = self.etree.Element
        SubElement = self.etree.SubElement

        a = Element('a')
        b = SubElement(a, 'b')
        c = SubElement(a, 'c')
        d = SubElement(a, 'd')
        e = SubElement(a, 'e')

        del a[::-1]
        self.assertEquals(
            [],
            list(a))

    def test_delslice_step_negative2(self):
        Element = self.etree.Element
        SubElement = self.etree.SubElement

        a = Element('a')
        b = SubElement(a, 'b')
        c = SubElement(a, 'c')
        d = SubElement(a, 'd')
        e = SubElement(a, 'e')

        del a[::-2]
        self.assertEquals(
            [b, d],
            list(a))

The error messages go like this:

    del a[1::2]
    ValueError: attempt to assign sequence of size 0 to extended slice of size 2
    del a[::-1]
    ValueError: attempt to assign sequence of size 0 to extended slice of size 4
    del a[::-2]
    ValueError: attempt to assign sequence of size 0 to extended slice of size 2


Additionally, I get this error:

    self.assertNotEquals(None, e.code)
    AttributeError: 'ParseError' object has no attribute 'code'

for this test:

    required_versions_ET['test_feed_parser_error_position'] = (1,3)
    def test_feed_parser_error_position(self):
        ParseError = self.etree.ParseError
        parser = self.etree.XMLParser()
        try:
            parser.close()
        except ParseError:
            e = sys.exc_info()[1]
            self.assertNotEquals(None, e.code)
            self.assertNotEquals(0, e.code)
            self.assert_(isinstance(e.position, tuple))
            self.assert_(e.position >= (0, 0))


The complete test suite is here:

https://github.com/lxml/lxml/blob/master/src/lxml/tests/test_elementtree.py

I can confirm that this indeed fails for the C implementation, while succeeding for the Python implementation.

The C implementation doesn't appear to support extended slices for getting and setting items. Looking into it.

Correction: extended slices are supported. The problem appears to be just with 'del', because element_ass_subscr doesn't treat it specially and thinks we just want to assign a 0-len value to a non-0-len slice.

Attaching a patch that fixes the slice deletion problems.

The approach I've taken is follow a similar implementation in Objects/listobject.c; when a slice is deleted, it can be done efficiently by using memmove to shift whole blocks of leftover children.

Also added a new test class to test_xml_etree specifically to test slicing with Element objects and deleting slices with strange steps.

I'll leave it up for review for a couple of days before committing.

WRT the ParseError problem, I've opened issue #14207 for it since the problems are quite unrelated.

New changeset 1a721b9a4039 by Eli Bendersky in branch 'default':
Issue #14178: Problem deleting slices with steps != +1 in the _elementtree module.
http://hg.python.org/cpython/rev/1a721b9a4039