[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-20 Thread Martin v . Löwis
Martin v. Löwis mar...@v.loewis.de added the comment: Amaury, I fail to see how the error you get on Windows is related to this issue. AFAICT, neither the issue nor the patch affects Windows at all. Closing the issue as fixed. If you think there is also an issue on Windows, please submit a

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread Amaury Forgeot d'Arc
Amaury Forgeot d'Arc amaur...@gmail.com added the comment: It does not work on Windows: subprocess.Popen(c:/windows/notepad.exe, cwd=b'c:/temp') Traceback (most recent call last): File stdin, line 1, in module File D:\afa\python\py3k-1\lib\subprocess.py, line 681, in __init__

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: And please add unit tests... I'm thinking on this. I plan to write tests for all my last changes about surrogates. -- ___ Python tracker rep...@bugs.python.org

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: It does not work on Windows I always consider Windows as a special case because Windows uses unicode internally. Byte string are converted quickly to unicode using the current locale. My patch was for UNIX/BSD which uses byte

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread Amaury Forgeot d'Arc
Amaury Forgeot d'Arc amaur...@gmail.com added the comment: PEP 277 explicitly states that unicode strings should be passed to wide-character functions, whereas byte strings use standard functions. This is done in posixmodule.c, for example. The current locale is a moving thing. --

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: PEP 277 explicitly states that unicode strings should be passed to wide-character functions, whereas byte strings use standard functions. This is done in posixmodule.c, for example. CreateProcessW takes a lot of arguments.

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-19 Thread Amaury Forgeot d'Arc
Amaury Forgeot d'Arc amaur...@gmail.com added the comment: mbcs is not a fixed encoding and may change between Windows sessions, see the Rationale in PEP277 http://www.python.org/dev/peps/pep-0277/ The mixed case is interesting. We could use CreateProcessW when at least one string is Unicode,

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-16 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: Commited in py3k (r80135), blocked in 3.1 (r80136). Python 3.1 has no _posixsubprocess module, it uses os.chdir() which accepts bytes, bytearray and str with surrogates. -- resolution: - fixed status: open - closed

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-15 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: New patch supporting bytearray type (remove the buggy assertion). -- Added file: http://bugs.python.org/file16939/posixsubprocess_cwd-2.patch ___ Python tracker rep...@bugs.python.org

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-13 Thread STINNER Victor
New submission from STINNER Victor victor.stin...@haypocalc.com: In py3k, subprocess uses _posixsubprocess.fork_exec() function. This function uses surrogateescape error handler for most arguments, but not for the current working directory (cwd). Attached patch uses PyUnicode_FSConverter() as

[issue8393] subprocess: support undecodable current working directory on POSIX OS

2010-04-13 Thread STINNER Victor
STINNER Victor victor.stin...@haypocalc.com added the comment: See also #8391. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue8393 ___ ___