If we are writing xml with CDATA section and leaving non-empty indentation and new-line parameters, a parent node of the section will contain useless indentation, that will be parsed as a text.

Example:
>>>doc = minidom.Document()
>>>root = doc.createElement('root')
>>>doc.appendChild(root)
>>>node = doc.createElement('node')
>>>root.appendChild(node)
>>>data = doc.createCDATASection('</data>')
>>>node.appendChild(data)
>>>print(doc.toprettyxml(indent=‘  ‘ * 4)
<?xml version="1.0" ?>
<root>
    <node>
<![CDATA[</data>]]>    </node>
</root>

If we try to parse this output doc, we won’t get CDATA value correctly.

Following code returns a string that contains only indentation characters:
>>>doc = minidom.parseString(xml_text)
>>>doc.getElementsByTagName('node')[0].firstChild.nodeValue

Returns a string with CDATA value and indentation characters:
>>>doc.getElementsByTagName('node')[0].firstChild.wholeText


But we have a workaround:
>>>data.nodeType = data.TEXT_NODE
…
>>>print(doc.toprettyxml(indent=‘  ‘ * 4)
<?xml version="1.0" ?>
<root>
    <node><![CDATA[</data>]]></node>
</root>

It will be parsed correctly:
>>>doc.getElementsByTagName('node')[0].firstChild.nodeValue
</data>

But I think it will be better if we fix the writing function, which would set this as default behavior.

Yes, this case is incorrect. Pretty printing should not change character content inside of a simple tag.

The PR looks good to me.

New changeset 384b81d923addd52125e94470b11d2574ca266a9 by Serhiy Storchaka (Vladimir Surjaninov) in branch 'master':
bpo-36407: Fix writing indentations of CDATA section (xml.dom.minidom). (GH-12514)
https://github.com/python/cpython/commit/384b81d923addd52125e94470b11d2574ca266a9

Should we backport this change? I am not sure.

I don't think this should be backported. Pretty-printing is not a production relevant feature, more of a "debugging, diffing and help users see what they get" kind of feature. It's good to have it fixed for the future, but we shouldn't bother users with it during a point release.