I have a crash in Python 3.x environments and don't have in 2.7.

===

Crashes in:

Python 3.7.4, Anaconda, Linux 64bit, lxml 4.3.3
OR
Python 3.7.4, Anaconda, Linux 64bit, lxml 4.4.1
OR
Python 3.6.5, Gentoo Linux 64bit, lxml 4.3.3

test.py:

import lxml
import lxml.etree

with open('test.xml', 'rb') as f:
    list(
        lxml.etree.iterparse(f)
    )

    # Traceback (most recent call last):
    # File "test.py", line 18, in <module>
    # lxml.etree.iterparse(f)
    # File "src/lxml/iterparse.pxi", line 209, in lxml.etree.iterparse.__next__
    # File "src/lxml/iterparse.pxi", line 194, in lxml.etree.iterparse.__next__
    # File "src/lxml/iterparse.pxi", line 225, in lxml.etree.iterparse._read_more_events
    # File "src/lxml/parser.pxi", line 1380, in lxml.etree._FeedParser.close
    # Segmentation fault
    list(
        lxml.etree.iterparse(f)
    )

test.xml:

<?xml version="1.0" encoding="UTF-8"?>
<root></root>

(or any else)

GDB's backtrace see in attachment.

===

Doesn't crash in:

Python 2.7.15, Gentoo Linux 64bit, lxml 4.3.3

# Traceback (most recent call last):
#   File "test.py", line 19, in <module>
#     lxml.etree.iterparse(f)
#   File "src/lxml/iterparse.pxi", line 209, in lxml.etree.iterparse.__next__
#    File "/home/sasha/_lxml-bug/test.xml", line 0
lxml.etree.XMLSyntaxError: no element found

===

See also: https://bugs.launchpad.net/lxml/+bug/1833050

This tracker is for bugs in CPython (the C implementation of Python). lxml is not a part of the standard library. Please use a proper tracker for reporting a bug in lxml.

Yes, I do. Please read all.

1) According to https://bugs.launchpad.net/lxml/+bug/1833050 (by lxml author's opinion) it's not a lxml bug.

2) I wouldn't said that but according to backtrace we have crash at Python-time.

3) Moreover crash existence depends on Python version!

Thanks.

Indeed, there is a bug in Python. It can be reproduced without lxml:

$ ./python -c "raise SyntaxError('error', (b'file', 1, 2, 'text'))"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
python: Objects/unicodeobject.c:397: _PyUnicode_CheckConsistency: Assertion `PyUnicode_Check(op)' failed.
Aborted (core dumped)

It has been fixed in 3.7+ by issue37467, but the fix was not backported to 3.6.

I understand that this bug is a duplicate of bpo-37467 that I fixed recently in 3.7, 3.8 and master branches:

commit f9b7457bd7f438263e0d2dd1f70589ad56a2585e
Author: Victor Stinner <vstinner@redhat.com>
Date:   Mon Jul 1 16:51:18 2019 +0200

    bpo-37467: Fix PyErr_Display() for bytes filename (GH-14504)


> It has been fixed in 3.7+ by issue37467, but the fix was not backported to 3.6.

Python 3.6 no longer accept bug fixes, only security fixes:
https://devguide.python.org/#status-of-python-branches

And this bug doesn't look like a security issue.

Python 2.7 doesn't seem to be affected by this issue.

In short, nothing can be one on the Python side: please upgrade to Python 3.7.

I close this issue as a duplicate of bpo-37467.

Ok, thanks!

Good that I was too lazy to open bug in June because our colleague did it better then me :)

Seems like the fix doesn't present in current 3.7 version (3.7.4). Will try in 3.7.5 and will close

UPD: Oh, closed? Ok. But *no*, it's not fixed in current 3.7 version (in master - yes). I wrote: 3.7.4. Let's check each other instead of... Sorry. Thanks.

https://github.com/python/cpython/commit/8cbffc4d96d1da0fbc38da6f34f2da30c5ffd601 has been merged after 3.7.4 has been released, correct. I will be part of the next 3.7.5 release.

About the workflow: we close a bug once the change is merged into the development branch, not when a release is published.

For the current 3.7.5 schedule, see:
https://www.python.org/dev/peps/pep-0537/#id4

In the meanwhile, avoid calling lxml.etree.iterparse(f) twice :-)