tarfile.py: fix GNU and USTAR formats to properly handle paths with special characters that are encoded with more than one byte each

[RoddyShuler]

BPO	24838
Nosy	@gustaebel, @vstinner, @serhiy-storchaka
Files	fix-tarfile-path-truncation.patch: Patch to fix tarfile truncation with multi-byte characters

^{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 = 'https://github.com/gustaebel'
closed_at = <Date 2016-04-19.10:02:47.483>
created_at = <Date 2015-08-10.18:04:23.355>
labels = ['type-bug', 'library']
title = 'tarfile.py: fix GNU and USTAR formats to properly handle paths with special characters that are encoded with more than one byte each'
updated_at = <Date 2016-11-29.12:04:09.406>
user = 'https://bugs.python.org/RoddyShuler'

bugs.python.org fields:

activity = <Date 2016-11-29.12:04:09.406>
actor = 'vstinner'
assignee = 'lars.gustaebel'
closed = True
closed_date = <Date 2016-04-19.10:02:47.483>
closer = 'lars.gustaebel'
components = ['Library (Lib)']
creation = <Date 2015-08-10.18:04:23.355>
creator = 'Roddy Shuler'
dependencies = []
files = ['40157']
hgrepos = []
issue_num = 24838
keywords = ['patch']
message_count = 7.0
messages = ['248363', '248576', '263713', '263719', '263722', '263723', '281986']
nosy_count = 5.0
nosy_names = ['lars.gustaebel', 'vstinner', 'python-dev', 'serhiy.storchaka', 'Roddy Shuler']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue24838'
versions = ['Python 3.5', 'Python 3.6']

[RoddyShuler]

GNU and USTAR formats use a special case if the file path is longer than 100 bytes. The detection for this, though, incorrectly checked for 100 characters rather than 100 bytes. So, if the length was close to but not exceeding 100 characters and included special characters such that the encoded length is greater than 100 bytes, the encoded string was truncated to 100 bytes and thus the resulting file name was truncated within the tar file.

For example...

/gt-education/Colección Educativa Guatemala/thumbs/Libro de Texto Comunicacion y Lenguaje 1 Grado.jpg

is truncated as:

/gt-education/Colección Educativa Guatemala/thumbs/Libro de Texto Comunicacion y Lenguaje 1 Grado.jp

The attached patch fixes this. Initially found on Python 3.3. Patch is tested on Linux with version 3.4.3-6 from Debian. Looking at the source code, I am pretty confident that the problem still exists upstream in Python 3.5.

[gustaebel]

Thanks for the detailed report and the patch. I haven't checked yet, but I suppose that the entire 3.x branch is affected. The first thing I have to do now is to come up with a comprehensive testcase.

[python-dev]

New changeset d08d6b776694 by Lars Gustäbel in branch '3.5':
Issue bpo-24838: tarfile's ustar and gnu formats now correctly calculate name and
https://hg.python.org/cpython/rev/d08d6b776694

New changeset e281a57d5b29 by Lars Gustäbel in branch 'default':
Issue bpo-24838: Merge tarfile fix from 3.5.
https://hg.python.org/cpython/rev/e281a57d5b29

[vstinner]

Tests fail on FreeBSD:

http://buildbot.python.org/all/builders/AMD64%20FreeBSD%209.x%203.5/builds/713/steps/test/logs/stdio

Example:

======================================================================
FAIL: test_unicode_link1 (test.test_tarfile.UstarUnicodeTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/usr/home/buildbot/python/3.5.koobs-freebsd9/build/Lib/test/test_tarfile.py", line 1807, in test_unicode_link1
    self._test_ustar_link("0123456789" * 9 + "01234567\xff")
  File "/usr/home/buildbot/python/3.5.koobs-freebsd9/build/Lib/test/test_tarfile.py", line 1826, in _test_ustar_link
    self.assertEqual(name, t.linkname)
AssertionError: '0123[44 chars]89012345678901234567890123456789012345678901234567\xff' != '0123[44 chars]89012345678901234567890123456789012345678901234567\udcc3\udcbf'
- 01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567\xff
?                                                                                                   ^
+ 01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567\udcc3\udcbf
?                                                                                                   ^^

[python-dev]

New changeset 78ede2baa146 by Lars Gustäbel in branch '3.5':
Issue bpo-24838: Fix test_tarfile.py for non-utf8 filesystem encodings.
https://hg.python.org/cpython/rev/78ede2baa146

New changeset 08835d1e7a50 by Lars Gustäbel in branch 'default':
Issue bpo-24838: Merge test_tarfile.py fix from 3.5.
https://hg.python.org/cpython/rev/08835d1e7a50

[gustaebel]

Sorry for the glitch, I suppose everything works fine now.

[vstinner]

FYI the first release including the fix 78ede2baa146 is Python 3.5.2.