[SciPy-Dev] [SciPy-User] ANN: SciPy 0.10.1 release candidate 1

Samuel John scipy at samueljohn.de
Tue Feb 14 16:54:00 EST 2012


[sorry posting to scipy-users and scipy-dev. I feel this is not a good idea, but this thread is already spanning both lists. Replay on scipy-dev pls]

The good news is that using clang and clang++ and gfortran from http://r.research.att.com/tools/ (4.2.4-5666.3) numpy and scipy build and test() fine! 
Yeeha \o/
Anyone with deeper understanding of the scipy internals want to comment on clang usage? Has anyone experienced problems with scipy and clang?

However, building numpy 1.6.1 and scipy 0.10.1rc1 (or 0.10.0 or head) on OS X 10.7.3 with Xcode 4.2.1 (build 4D502) with llvm-gcc (which is the default, since non-llvm gcc is not available any more with XCode 4.2) leads to a segfault or to a malloc trap:

> Running unit tests for scipy
> NumPy version 1.6.1
> NumPy is installed in /usr/local/Cellar/python/2.7.2/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/numpy
> SciPy version 0.10.1rc1
> SciPy is installed in /usr/local/Cellar/python/2.7.2/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/scipy
> Python version 2.7.2 (default, Feb 14 2012, 22:09:10) [GCC 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.1.00)]
> nose version 1.1.2
> ...................................................................................................................................................................................F.FFFPython(13536,0x7fff7c026960) malloc: *** error for object 0x10740b368: incorrect checksum for freed object - object was probably modified after being freed.
> *** set a breakpoint in malloc_error_break to debug
> Abort trap: 6


As pip install will also use the default llvm-gcc this might be a severe issue! This is already the case right now but does not often show up in daily usage.
Perhaps it's possible to set the shell vars CC and CXX during build?

Concerning arpack: I'm afraid that the arpack issues prevail in 0.10.1rc1. I beg you to fix these (I am not able to). 
scipy.test() log at https://gist.github.com/1830780

Samuel





More information about the SciPy-Dev mailing list