Segmentation fault in TRAIN step

[hdetering]

Hi Andrew,

using JointSNVMix v0.8, jsm.py train gives me the following error:

$ jsm.py train hs37d5.fa CS.H.cigar19.bam CS.T.cigar19.bam trained_params.txt
Segmentation fault

Input

The input files used are attached:
CS.cigar19.zip

As far as could narrow down the problem, it occurs when the CIGAR string exceeds a length of 18 characters. With the following input, the program ran without errors:
CS.cigar18.zip

Could the fix be as easy as updating the samtools sources?

Additional information:

I installed JointSNVMix inside a Miniconda container, with Python 2.7.15:

wget https://github.com/aroth85/JointSNVMix/raw/master/dist/JointSNVMix-0.8.tar.gz
tar xf JointSNVMix-0.8.tar.gz
cd JointSNVMix-0.8
conda create --prefix $PWD/env python=2.7
source activate ./env
python setup.py install

We are using the human reference genome, downloaded from this location: ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/phase2_reference_assembly_sequence

Thanks and kind regards.
Harry