distutils does not show any error msg when can't build C module extensions due to a missing C compiler

[giampaolo]

BPO	4931
Nosy	@loewis, @amauryfa, @giampaolo, @tarekziade, @merwok, @takluyver
Files	distutils_ioerror.patch

^{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/merwok'
closed_at = <Date 2014-03-12.08:25:30.156>
created_at = <Date 2009-01-13.17:31:04.888>
labels = ['type-bug', 'library']
title = "distutils does not show any error msg when can't build C module extensions due to a missing C compiler"
updated_at = <Date 2014-03-12.08:25:30.154>
user = 'https://github.com/giampaolo'

bugs.python.org fields:

activity = <Date 2014-03-12.08:25:30.154>
actor = 'eric.araujo'
assignee = 'eric.araujo'
closed = True
closed_date = <Date 2014-03-12.08:25:30.156>
closer = 'eric.araujo'
components = ['Distutils', 'Distutils2']
creation = <Date 2009-01-13.17:31:04.888>
creator = 'giampaolo.rodola'
dependencies = []
files = ['12739']
hgrepos = []
issue_num = 4931
keywords = ['patch']
message_count = 15.0
messages = ['79751', '79767', '79769', '79845', '101389', '101390', '117690', '120588', '120611', '212907', '212909', '213207', '213208', '213213', '213219']
nosy_count = 8.0
nosy_names = ['loewis', 'jafo', 'amaury.forgeotdarc', 'giampaolo.rodola', 'tarek', 'eric.araujo', 'python-dev', 'takluyver']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue4931'
versions = ['Python 2.7', 'Python 3.3', 'Python 3.4']

[giampaolo]

Today I was trying to compile a module using an extension in C and
noticed there are differences between compiling it on Python 2.5 and 2.6.
I was trying to compile psutil (svn checkout
http://psutil.googlecode.com/svn/trunk/ psutil) but I think that the
problem may occur with any other package using C extensions.

Python 2.5:

D:\pyftpdlib\svn\psutil\trunk>C:\python25\python.exe setup.py install
running install
running build
running build_ext
error: Python was built with Visual Studio 2003;
extensions must be built with a compiler than can generate compatible
binaries.
Visual Studio 2003 was not found on this system. If you have Cygwin
installed,
you can try compiling with MingW32, by passing "-c mingw32" to setup.py.

Python 2.6:

D:\pyftpdlib\svn\psutil\trunk>C:\python26\python.exe setup.py install
running install
running build
running build_ext
building 'psutil/_psutil_mswindows' extension
error: None

[loewis]

What is the specific error you are reporting?

[giampaolo]

Currently I haven't any C compiler installed on my system so I expect
distutils to report that every time I try to compile a C module extension.
Python 2.5 did that while Python 2.6 does not.

[amauryfa]

The problem has two causes:

• IOError is raised when no compiler is found. DistutilsPlatformError
  should be used instead, this was fixed by bpo-4702.

• the distutils.util.grok_environment_error function transforms a
  IOError("Unable to find vcvarsall.bat") into "error: None".

This function should not be used IMO. Its docstring claims that it
"Handles Python 1.5.1 and 1.5.2 styles", but str() does a better job...

Patch attached.

[jafo]

Tarek: This patch seems reasonable to me, is this something that can be applied?

[tarekziade]

I guess i can be applied on distutils, and backported in distutils2. I'll do it in the coming days.

Notice that I am now applying only bug fixes and regression fixes now for distutils.

[merwok]

I will commit this if Tarek does not do it shortly.

In distutils2, I’ll remove grok_environment_error wholly, since it exists for 1.5 compat only.

[merwok]

Giampaolo, can you test Amaury’s patch? I’d prefer some testing before committing.

[giampaolo]

I don't have the same setup I had at the time when I submitted the first message so maybe something has changed in meantime.
Below is what I get on Windows 2000 SP-3, no VS installed, python 2.7, before and after the patch.

C:\Documents and Settings\foo\Desktop\psutil>setup.py build
running build
running build_py
copying psutil\compat.py -> build\lib.win32-2.7\psutil
copying psutil\error.py -> build\lib.win32-2.7\psutil
copying psutil\_psbsd.py -> build\lib.win32-2.7\psutil
copying psutil\_pslinux.py -> build\lib.win32-2.7\psutil
copying psutil\_psmswindows.py -> build\lib.win32-2.7\psutil
copying psutil\_psosx.py -> build\lib.win32-2.7\psutil
copying psutil\_psposix.py -> build\lib.win32-2.7\psutil
copying psutil\init.py -> build\lib.win32-2.7\psutil
running build_ext
building '_psutil_mswindows' extension
error: Unable to find vcvarsall.bat

Although the message is improved ("Unable to find vcvarsall.bat" vs "None") it seems the patch had no effect.

[takluyver]

Any chance of getting this patch applied? It clearly makes the error message more useful, and we've run into another case where grok_environment_error gives the wrong result: when symlinking fails because the target exists, it now says "File exists: <source>", because e.filename is the source and e.filename2 the target.

It's also rather embarassing that a function in Python 3.4 still says "Handles Python 1.5.1 and 1.5.2 styles..." in the docstring.