XML special character in the context can result in corrupted DOCX

[keestux]

Describe the bug

Special XML characters in "the context" can result in corrupted DOCX output.

To Reproduce

docx example would contain

{%p for  defect  in defects %}
Criticality: {{ defect.criticality }}
Source: {{ defect.source }}
{%p endfor %}

The context in json format

{
  "defects": [
    {
      "criticality": "<select>",
      "source": "QA"
    },
    {
      "criticality": "Safety Related",
      "source": "QA"
    }
  ]
}

No error is given during the template generation, but the .docx is corrupt. You can still open the .docx, but
the contents looks like this:

Criticality: 

Expected behavior

The generated .docx should be like this:

Safety Criticality: <select>
Source: QA

Safety Criticality: Safety Related
Source: QA

[keestux]

    def build_xml(self, context, jinja_env=None):
        xml = self.get_xml()
        xml = self.patch_xml(xml)
        xml = self.render_xml_part(xml, self.docx._part, context, jinja_env)

When you stop the debugger, right before doing self.render_xml_part you can check the XML validity in the debugger console

etree.fromstring(xml)

You'll get something like

<Element {http://schemas.openxmlformats.org/wordprocessingml/2006/main}body at 0x7fc381147c40>

Next, you step over the render_xml_part and look at the xml again, you'll notice an XML error

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "src/lxml/etree.pyx", line 3428, in lxml.etree.fromstring
  File "src/lxml/parser.pxi", line 2066, in lxml.etree._parseMemoryDocument
  File "src/lxml/parser.pxi", line 1921, in lxml.etree._parseDoc
  File "src/lxml/parser.pxi", line 1937, in lxml.etree._parseDoc_unicode
  File "src/lxml/parser.pxi", line 1136, in lxml.etree._BaseParser._parseUnicodeDoc
  File "src/lxml/parser.pxi", line 647, in lxml.etree._ParserContext._handleParseResultDoc
  File "src/lxml/parser.pxi", line 765, in lxml.etree._handleParseResult
  File "src/lxml/parser.pxi", line 689, in lxml.etree._raiseParseError
  File "<string>", line 1
lxml.etree.XMLSyntaxError: Opening and ending tag mismatch: select line 1 and w:t, line 1, column 1626

Later, the XML is somewhat sanitized in fix_tables

    def fix_tables(self, xml):
        parser = etree.XMLParser(recover=True)
        tree = etree.fromstring(xml, parser=parser)

Now the tree is valid XML. The offending <select> is converted into <select/>. But the resulting .docx is not OK. The text in the document stops at the spot where <select/> is.

[keestux]

As a (temporary?) workaround you can sanitize the context, escaping special XML characters, as described here

For example with this function:

xml_char_table = str.maketrans({
    "<": "&lt;",
    ">": "&gt;",
    "&": "&amp;",
    "'": "&apos;",
    '"': "&quot;",
})
def xmlesc(txt):
    return txt.translate(xml_char_table)