GzipFile doesn't always ignore None as filename

[da]

BPO	33038
Nosy	@benjaminp, @ned-deily, @serhiy-storchaka, @bbayles
PRs	[2.7] bpo-33038: gzip.GzipFile assumes non-None name attribute #6095
Files	gzip_filename_fix.patch: Diff for suggested fix

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2018-05-09.10:19:46.430>
created_at = <Date 2018-03-10.00:46:10.803>
labels = ['library', 'type-crash']
title = "GzipFile doesn't always ignore None as filename"
updated_at = <Date 2018-05-09.10:19:46.430>
user = 'https://bugs.python.org/da'

bugs.python.org fields:

activity = <Date 2018-05-09.10:19:46.430>
actor = 'serhiy.storchaka'
assignee = 'none'
closed = True
closed_date = <Date 2018-05-09.10:19:46.430>
closer = 'serhiy.storchaka'
components = ['Library (Lib)']
creation = <Date 2018-03-10.00:46:10.803>
creator = 'da'
dependencies = []
files = ['47473']
hgrepos = []
issue_num = 33038
keywords = ['patch']
message_count = 8.0
messages = ['313512', '313578', '313632', '316010', '316011', '316091', '316112', '316323']
nosy_count = 5.0
nosy_names = ['benjamin.peterson', 'ned.deily', 'serhiy.storchaka', 'bbayles', 'da']
pr_nums = ['6095']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue33038'
versions = ['Python 2.7']

[da]

The Python documentation states that if the GzipFile can't determine a filename from fileobj it'll use an empty string and won't be included in the header. Unfortunately, this doesn't work for SpooledTemporaryFile which has a name attribute but doesn't set it initially. The result is a crash.

To reproduce

import gzip
import tempfile

with tempfile.SpooledTemporaryFile() as fd:
    with gzip.GzipFile(mode='wb', fileobj=fd) as gz:
        gz.write(b'asdf')

Result:

Traceback (most recent call last):
  File "<stdin>", line 2, in <module>
  File "/Users/diegoargueta/.pyenv/versions/2.7.14/lib/python2.7/gzip.py", line 136, in __init__
    self._write_gzip_header()
  File "/Users/diegoargueta/.pyenv/versions/2.7.14/lib/python2.7/gzip.py", line 170, in _write_gzip_header
    fname = os.path.basename(self.name)
  File "/Users/diegoargueta/.pyenv/versions/gds27/lib/python2.7/posixpath.py", line 114, in basename
    i = p.rfind('/') + 1
AttributeError: 'NoneType' object has no attribute 'rfind'

This doesn't happen on Python 3.6, where the null filename is handled properly. I've attached a patch file that fixed the issue for me.

[bbayles]

da, would you mind if I add a test and a news entry to your patch and submit it as a Github pull request?

[da]

Yeah that's fine. Thanks!

[da]

Did this make it into 2.7.15? There aren't any release notes for it on the download page like usual.

[ned-deily]

It looks like PR 6095 for this issue has not been reviewed or merged yet.

The commits for v2.7.15 are here:
https://github.com/python/cpython/commits/v2.7.15

[bbayles]

Is there someone who might be in a position to review the PR? It's pretty short.

[serhiy-storchaka]

The workaround in the current 2.7 is specifying an explicit filename argument:

with tempfile.SpooledTemporaryFile() as fd:
    with gzip.GzipFile(filename='', mode='wb', fileobj=fd) as gz:
        gz.write(b'asdf')

[serhiy-storchaka]

New changeset afe5f63 by Serhiy Storchaka (Bo Bayles) in branch '2.7':
bpo-33038: Fix gzip.GzipFile for file objects with a non-string name attribute. (GH-6095)
afe5f63