This function function often result in incorrect output when the cpu is very loaded. I do not know how to trace the bug since every "single shot" use, or step by step trace gives the correct answer, also when running scripts under "ipython -pdb" solves the problem.
Python 2.6.2 (64 bits)
numpy/scipy are built with atlas support.
OS: Gentoo linux (I use two independent (not a clone of each other) machines).
Sorry for posting it on the numpy list, but anyway this is not the first cross-lists post.
It seems that repr applied do an object array does not provide the info
needed to recreate it:
In : y = array([Decimal(1)]*2)
In : repr(y)
Out: 'array([1, 1], dtype=object)'
And of course, there is going to be a problem with arrays of more than one
dimension anyway. But I wonder if this should be fixed?