Multi-allelic and complex variants lose zygosity info when atomized

[marbeno]

Hello,
using version 1.20, I have a bunch of variants that might get the wrong genotype when normalized. You can finde the test file attached
test.vcf.gz

For example (INFO e FORMAT removed for readibility):

chr14	105174100	.	GCCCGC	CGCCCCGC,GCCCG	18.6776	PASS	.	.	0/1

Becomes

bcftools norm -a --atom-overlaps . test.vcf.gz | bcftools norm -f hg19.fasta
chr14	105174100	.	G	C	18.6776	PASS	.	.	0/.
chr14	105174100	.	G	GGC	18.6776	PASS	.	.	0/.
chr14	105174104	.	GC	G	18.6776	PASS	.	.	0/0

While a I would expect something like

chr14	105174100	.	G	C	18.6776	PASS	.	.	0/.
chr14	105174100	.	G	GGC	18.6776	PASS	.	.	0/1
chr14	105174104	.	GC	G	18.6776	PASS	.	.	0/0

The second record:

chr22	36744886	.	GCCCC	GGCT	578.003	PASS	.	.	0/1

Becomes

bcftools norm -a --atom-overlaps . test.vcf.gz | bcftools norm -f hg19.fasta
chr22	36744886	.	GC	G	578.003	PASS	.	.	0/.
chr22	36744887	.	C	G	578.003	PASS	.	.	0/1
chr22	36744889	.	C	T	578.003	PASS	.	.	0/.

while I would expect something like

chr22	36744886	.	GC	G	578.003	PASS	.	.	0/1
chr22	36744888	.	C	G	578.003	PASS	.	.	0/1
chr22	36744890	.	C	T	578.003	PASS	.	.	0/1

[pd3]

Thank you for raising the issue and the test case. This is now fixed.