Erroneous tree parsed

[Jason R. Coombs]

Using 4.4.1 on Python 3.5.0 on Linux with a late build of lxml, I run this script:

"""
This script outputs:
[<roo/>]
roo

[<root/>]
root

[]

"""

import io
import bs4

doc = """<?xml version="1.0" encoding="us-ascii"?>
<root />
"""

def report(soup):
 elems = soup.findAll()
 print(elems)
 if elems:
  print(elems[0].name)
 print()

report(bs4.BeautifulSoup(io.StringIO(doc), 'xml'))
report(bs4.BeautifulSoup(io.StringIO(doc.replace('us-ascii', 'utf-8')), 'xml'))

doc2 = """<?xml version="1.0" encoding="us-ascii"?>
<?OFX OFXHEADER="200" VERSION="200"?>
<root />
"""
report(bs4.BeautifulSoup(io.StringIO(doc2), 'xml'))

---

as you can see, depending on the 'encoding' declaration in the header, the sole root tag might be corrupted. Furthermore, adding a second prolog causes the document to parse as empty.

Parsing these same documents with lxml directly works just fine.

[Jason R. Coombs (jaraco)]

With Python 3.5.1 on OS X, I no longer see the error. I suspect the issue was corrected in Python 3.5.1.

[Leonard Richardson (leonardr)]

Wow, that's really weird. Since you're reporting that it's been fixed I'm going to close this bug, but I'm not sure about the fix happening in Python. This has symptoms like a lot of lxml bugs I've seen.