[Numpy-tickets] [NumPy] #619: remove split_quoted in numpy.distutils.ccompiler

NumPy numpy-tickets@scipy....
Thu Apr 10 05:49:41 CDT 2008


#619: remove split_quoted in numpy.distutils.ccompiler
-----------------------------+----------------------------------------------
 Reporter:  jarrod.millman   |        Owner:  jarrod.millman
     Type:  task             |       Status:  reopened      
 Priority:  normal           |    Milestone:  1.0.5         
Component:  numpy.distutils  |      Version:  devel         
 Severity:  normal           |   Resolution:                
 Keywords:                   |  
-----------------------------+----------------------------------------------
Changes (by pearu):

  * status:  closed => reopened
  * resolution:  fixed =>

Comment:

 I disagree with this change. The split_quoted was introduced to
 support windows platform where there exist very subtle and unlocal
 issues with the path names with spaces that are not easy to reproduce
 on other platforms. For example, I believe that the issue 723
 was triggered by removing split_quoted from numpy.distutils.ccompiler.
 Some of these issues may be fixed in std distutils but we still
 need to support Python 2.3 that do not contain the fixes.

 Regarding the clean up of numpy.distutils, my first suggestion is,
 don't do it, you may break it for other platforms. Don't do it
 even if you or somebody else thinks that it should be harmless
 At least, ask the author of a particular code if it is safe to clean
 up - distutils just is a very sensitive piece of software.

 My second suggestion is, if you must do it, make sure to test
 the changes with all possible combinations of platforms,
 compilers, and possible installation paths.
 On windows, it basically means testing with MSVC, Mingw, cygwin, gfortran,
 Intel
 compilers that are installed using paths with and without spaces.
 This can be a very tedious task. If you cannot take it, use my first
 suggestion.

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/619#comment:5>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.


More information about the Numpy-tickets mailing list