open.source/magic-wormhole
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

add notes, make test TODO instead of SKIP

Browse Source
This commit is contained in:
Brian Warner 2017-11-01 17:42:51 -07:00
parent d727531e6d
commit 1c5f29337e
2 changed files with 24 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/wormhole/cli/cmd_send.py
View File

@ -254,18 +254,28 @@ class Sender:
# We use realpath() instead of normpath() to locate the actual
# file/directory, because the path might contain symlinks, and
# normpath() would collapse those before resolving them.
# test_cli.OfferData.test_symlink_collapse tests this.
# Unfortunately on windows, realpath() is built out of normpath()
# because of a no-longer-true belief that windows does not have a
# working os.path.islink(): see https://bugs.python.org/issue9949
# The consequence is that "wormhole send PATH" might send the wrong
# file, if:
# Unfortunately on windows, realpath() (on py3) is built out of
# normpath() because of a py2-era belief that windows lacks a working
# os.path.islink(): see https://bugs.python.org/issue9949 . The
# consequence is that "wormhole send PATH" might send the wrong file,
# if:
# * we're running on windows
# * PATH goes down through a symlink and then up with parent-directory
# navigation (".."), then back down again
# * the back-down-again portion of the path also exists under the
# original directory (an error is thrown if not)
# I'd like to fix this. The core issue is sending directories with a
# trailing slash: we need to 1: open the right directory, and 2: strip
# the right parent path out of the filenames we get from os.walk(). We
# used to use what.rstrip() for this, but bug #251 reported this
# failing on windows-with-bash. realpath() works in both those cases,
# but fails with the up-down symlinks situation. I think we'll need to
# find a third way to strip the trailing slash reliably in all
# environments.
what = os.path.realpath(what)
if not os.path.exists(what):
raise TransferError("Cannot send: no file/directory named '%s'" %

15
src/wormhole/test/test_cli.py
View File

@ -203,12 +203,6 @@ class OfferData(unittest.TestCase):
def test_symlink_collapse(self):
if not hasattr(os, 'symlink'):
raise unittest.SkipTest("host OS does not support symlinks")
if os.name == "nt":
# ntpath.py's realpath() is built out of normpath(), and does not
# follow symlinks properly, so this test always fails. "wormhole
# send PATH" on windows will do the wrong thing. See
# https://bugs.python.org/issue9949" for details.
raise unittest.SkipTest("host OS has broken os.path.realpath(), see https://bugs.python.org/issue9949")
# build A/B1, A/B1/D.txt
# A/B2/C2, A/B2/D.txt
# symlink A/B1/C1 -> A/B2/C2
@ -239,6 +233,15 @@ class OfferData(unittest.TestCase):
d, fd_to_send = build_offer(self.cfg)
self.assertEqual(d["file"]["filename"], "D.txt")
self.assertEqual(fd_to_send.read(), b"success")
if os.name == "nt":
test_symlink_collapse.todo = "host OS has broken os.path.realpath()"
# ntpath.py's realpath() is built out of normpath(), and does not
# follow symlinks properly, so this test always fails. "wormhole send
# PATH" on windows will do the wrong thing. See
# https://bugs.python.org/issue9949" for details. I'm making this a
# TODO instead of a SKIP because 1: this causes an observable
# misbehavior (albeit in rare circumstances), 2: it probably used to
# work (sometimes, but not in #251). See cmd_send.py for more notes.
class LocaleFinder:
def __init__(self):