Fix a bug in splitting Type=String FORMAT fields via -a and -m

Fixes #2476

Author: pd3

NEWS

@@ -12,6 +12,10 @@ Changes affecting specific commands:
 
     - Add new FORMAT/QM annotation, to be used with the new `bcftools +trio-dnm3 --use-ALM` model.
 
+* bcftools norm
+
+    - Fix a bug in splitting Type=String FORMAT fields via -a and -m (#2476)
+
 * bcftools +prune
 
     - When pruning by window length in base pairs, one out-of-window record would

abuf.c

@@ -1,6 +1,6 @@
 /* The MIT License
 
-   Copyright (c) 2021-2024 Genome Research Ltd.
+   Copyright (c) 2021-2026 Genome Research Ltd.
 
    Author: Petr Danecek <pd3@sanger.ac.uk>
 
@@ -630,8 +630,9 @@ static void _split_table_set_format(abuf_t *buf, bcf_fmt_t *fmt, merge_rule_t mo
             memset(buf->tmp2,0,nval1_dst*nsmpl);
             for (i=0; i<nsmpl; i++)
             {
+                char *src = (char*)buf->tmp2 + nval1_dst*i;
                 kstring_t dst;
-                dst.l = 0; dst.m = nval1_dst; dst.s = (char*)buf->tmp2 + nval1_dst*i;
+                dst.l = 0; dst.m = nval1_dst + 1; dst.s = src;
                 kputc_('.',&dst);
                 if ( star_allele ) kputsn_(",.",2,&dst);
                 if ( len==BCF_VL_R )
@@ -641,6 +642,7 @@ static void _split_table_set_format(abuf_t *buf, bcf_fmt_t *fmt, merge_rule_t mo
                 }
                 copy_string_field(buf->tmp+nval1*i, iori+ioff, nval1, &dst, 0+ioff);
                 if ( star_allele ) copy_string_field(".", 0, 1, &dst, 1+ioff);
+                assert( src==dst.s );   // make sure copy_string_field() did not reallocate
             }
             ret = bcf_update_format(buf->out_hdr, out, tag, buf->tmp2, nval1_dst*nsmpl, type);
         }

test/test.pl

@@ -304,6 +304,7 @@
 run_test(\&test_vcf_norm,$opts,in=>'norm.symbolic.3',out=>'norm.symbolic.3.1.out',fai=>'norm.symbolic.3',args=>'');
 run_test(\&test_vcf_norm,$opts,in=>'norm',out=>'norm.out',fai=>'norm',args=>'-cx');
 run_test(\&test_vcf_norm,$opts,in=>'norm.split',out=>'norm.split.out',args=>'-m-');
+run_test(\&test_vcf_norm,$opts,in=>'norm.split',out=>'norm.split.1.out',args=>'-m- -a');
 run_test(\&test_vcf_norm,$opts,in=>'norm.split.2',out=>'norm.split.2.out',args=>'-m-');
 run_test(\&test_vcf_norm,$opts,in=>'norm.split.3',out=>'norm.split.3.out',args=>'-m- --force');
 run_test(\&test_vcf_norm,$opts,in=>'norm.split.4',out=>'norm.split.4.1.out',args=>'-m-');