[NumPy-Tickets] [NumPy] #1724: new_iterator branch - bus error

NumPy Trac numpy-tickets@scipy....
Sat Jan 29 00:58:03 CST 2011

#1724: new_iterator branch - bus error
 Reporter:  rgommers                  |       Owner:  somebody   
     Type:  defect                    |      Status:  new        
 Priority:  normal                    |   Milestone:  Unscheduled
Component:  numpy.core                |     Version:  none       
 Keywords:  new_iterator, half-float  |  

Comment(by rgommers):

 I assume you mean whether or not I removed the build folder before a new
 build? Yes, I do:
 $ git clean -xdf
 $ python setup.py build_ext -i
 A build folder present or not after building that makes any difference is
 something I've never heard of. Just in case I've checked, both with and
 without it I get segfaults.

 What I missed in the build log the first time around is the following,
 right at the start:
 Running from numpy source directory.non-existing path in
 'numpy/distutils': 'site.cfg'
 F2PY Version 2
 numpy/core/setup_common.py:86: MismatchCAPIWarning: API mismatch detected,
 the C API
 version numbers have to be updated. Current C api version is 5, with
 72b9c4298262b6e0e3d4c77845f48135, but recorded checksum for C API version
 5 in
 codegen_dir/cversions.txt is 77e2e846db87f25d7cf99f9d812076f0. If
 functions were
 added in the C API, you have to update C_API_VERSION  in

 It's not just running the tests that gives a segfault, a simple
 >>> import numpy
 >>> numpy.linspace(1, 10)
 will do it too.

 I was asking before on the list for someone to review the ABI fixes
 because my understanding from 1.4.0 was that it is just not possible to
 add new types while keeping ABI compatibility, because the size of the
 struct holding the types increases. Disclaimer: I don't know too much
 about the numpy core, so I may have misunderstood.

 Another thing to fix is that any changes to setup.py should be applied to
 the corresponding SConscript file.

Ticket URL: <http://projects.scipy.org/numpy/ticket/1724#comment:6>
NumPy <http://projects.scipy.org/numpy>
My example project

More information about the NumPy-Tickets mailing list