Hi there,

I'm getting Segmentation fault (core dumped) errors when passing somewhat large matrices to solvers.qp(). The P matrix has shape (34638, 34638), and G matrix has shape (138552, 34638) (both sparse matrices). It doesn't seem to be a memory issue since the machine has 100GB of allocated memory and peak memory usage remained at ~30GB before the crash.

After googling a bit, I tried to reproduce this behavior on my local machine (with considerably less memory) and found that segmentation fault appears after the matrices exceeds a certain threshold, while under the threshold I get a memory error instead. This is very similar to what's described in https://groups.google.com/forum/#!topic/cvxopt/nxBXtv5WUV4.

In [1]: import cvxopt

In [2]: cvxopt.__version__
Out[2]: u'1.2.0'

In [3]: i = 32750

In [4]: G = cvxopt.spmatrix(1,(0,i*4),(0,i))

In [5]: G + 1
---------------------------------------------------------------------------
MemoryError                               Traceback (most recent call last)
<ipython-input-5-d684893d245a> in <module>()
----> 1 G + 1

MemoryError:

In [6]: i = 32780

In [7]: G = cvxopt.spmatrix(1,(0,i*4),(0,i))

In [8]: G + 1
Segmentation fault: 11

The original post mentioned that this is fixed in 1.1.9, and I'm using 1.2.0 here. Any pointers/thoughts on how I should go about debugging this? Thanks a lot in advance for your help!