bctools norm breaks in combination of `--atomize` and checking REF alleles

[thomasbtf]

When running bcftools with --atomize it results in a Segmentation fault (core dumped) :

> bcftools norm --output-type z -o out.vcf.gz  -f main.fasta --check-ref w --rm-dup exact --atomize in.vcf
Segmentation fault (core dumped)

However, when I remove the --atomize bcftools norm works:

> bcftools norm --output-type z -o out.vcf.gz  -f main.fasta --check-ref w --rm-dup exact in.vcf
NON_ACGTN_REF   NC_045512.2     28271   -
REF_MISMATCH    NC_045512.2     28280   C       G
REF_MISMATCH    NC_045512.2     28281   T       A
Lines   total/split/realigned/skipped:  550/0/0/0

When there are only matching REF alleles, it also seems to work fine.

The data from the above example is available here:
main.fasta
in.vcf

[pd3]

Thank you for the bug report.

The problem was caused by a faulty VCF violating the specification, the REF allele contained the dash - character for some reason). The code assumed that a valid VCF was being processed. A check was now added and the program no longer crashes on the provided test case.