Regression in unescaping backslash in attribute

[oprypin]

Regression in 3.4 - c0f6e5a is the first bad commit

Reproduction code:

import markdown

s = r'''
<img src="..\..\foo.png">
'''
print(repr(markdown.markdown(s)))

Output before vs after:

'<p><img src="..\\..\\foo.png"></p>'

'<p><img src="..\\\x0246\x03.\\foo.png"></p>'

[waylan]

This is definitely a bug. The final output should not still contain placeholders. Thank you for the report.

However, even if that is resolved, the input still has a problem. Consider the simple input \.. That would correctly output <p>.</p>. Notice that the backslash is removed; which is the correct behavior as the escape character should not be in the final output. However, in a URL ..\ has a specific meaning and the backslash should remain in the output. Therefore, the document author should escape the backslash: ..\\. Although, it appears that that is not currently working correctly either.

I will note that as a workaround, the document author can use forward slashes instead. The only system which uses backslashes is the DOS/Windows local file system, and most browsers will translate from forward slashes to backward slashes properly. Of course, that's not an excuse for the bug, but it does avoid the escaping issue altogether.

...and I just realized this issue is specific to raw HTML. We shouldn't be altering the raw HTML at all.

[oprypin]

Thanks for the quick fix