bcftools merge unexpected output

[biorococo]

I've found an unexpected output when merging gVCF files using bcftools v1.17.

I have 3 individuals with following gVCF block/variants:

SAMPLE1
chr20 64249835 . T <NON_REF> 0 . END=64249837 GT:GQ:MIN_DP:PL 0/0:50:54:0,162,1610
chr20 64249838 . T <NON_REF> 0 . END=64251648 GT:GQ:MIN_DP:PL 0/0:50:68:0,216,2150
SAMPLE2
chr20 64249835 . T <NON_REF> 0 . END=64250066 GT:GQ:MIN_DP:PL 0/0:50:38:0,114,1130
chr20 64250067 . T <NON_REF> 0 . END=64250067 GT:GQ:MIN_DP:PL 0/0:21:47:0,21,1160
SAMPLE3
chr20 64249835 . T <NON_REF> 0 . END=64249836 GT:GQ:MIN_DP:PL 0/0:50:28:0,84,830
chr20 64249837 . T <NON_REF> 0 . END=64250066 GT:GQ:MIN_DP:PL 0/0:50:38:0,117,1160

Using the following merge command:
bcftools merge SAMPLE_1_test.vcf.gz SAMPLE_2_test.vcf.gz SAMPLE_3_new.vcf.gz --gvcf reference.fasta

I get the following result:
chr20 64249835 . T <NON_REF> 0 . END=64249836 GT:GQ:MIN_DP:PL 0/0:50:54:0,162,1610 0/0:50:38:0,114,1130 0/0:50:28:0,84,830
chr20 64249837 . T <NON_REF> 0 . . GT:GQ:MIN_DP:PL ./.:.:.:. 0/0:50:38:0,114,1130 0/0:50:38:0,117,1160
chr20 64249838 . T <NON_REF> 0 . END=64250066 GT:GQ:MIN_DP:PL 0/0:50:68:0,216,2150 0/0:50:38:0,114,1130 0/0:50:38:0,117,1160

It looks like bcftools is failing to extract info from last position of gVCF block as position 64249837 is reported in all three samples but having "./.:.:.:." reported for SAMPLE_1 in the merge VCF output.

Any help on this would be appreciated. Thanks!

[pd3]

Thank you for the bug report and the test case. Hopefully this should be fixed by 2f482de, please try it out.

Just to cross-reference, this is related to #1891 and #1164