Error processing BAM file generated by bwa-mem2

[Hoeze]

Hi, we just got a segmentation fault when processing a BAM file generated by bwa-mem2:

# mosdepth --version
mosdepth 0.3.10
# mosdepth --threads 1 --by index0_germline1.converted_bed.bed --fasta GRCh38.fasta.gz index0_germline1 index0_germline1.md.bam                                               
SIGSEGV: Illegal storage access. (Attempt to read from nil?)
Segmentation fault (core dumped)

I tried updating to the latest version, and that indeed solves the SegFault. However, I still get an error message:

# mosdepth --version
mosdepth 0.3.12
# mosdepth --threads 1 --by index0_germline1.converted_bed.bed --fasta GRCh38.fasta.gz index0_germline1 index0_germline1.md.bam
fatal.nim(53)            sysFatal
Error: unhandled exception: value out of range: -149675 notin 0 .. 9223372036854775807 [RangeDefect]

What does that error mean? How do I solve that?

[brentp]

Hi, does your bed file have negative coordinates?
Is your bam file readable with samtools view ?

[Hoeze]

Hi @brentp, thanks for your quick response!

does your bed file have negative coordinates?

No, all coordinates are positive and start is strictly lower than end.
(Checked with awk '$2 < 0 || $3 < 0 || $2 >= $3' index0_germline1.converted_bed.bed)

Is your bam file readable with samtools view ?

Yes. Both samtools view and samtools quickcheck work.

[brentp]

mosdepth_debug.gz
Hi, can you run it with this binary and let me know the result? It should give us more info about where the error is occuring.

[Hoeze]

@brentp Thanks, here is the output:

# ./mosdepth_debug --threads 1 --by index0_germline1.converted_bed.bed --fasta GRCh38.fasta.gz index0_germline1 index0_germline1.md.bam
[mosdepth] WARNING: built in debug mode; will be slow
/home/brentp/src/mosdepth/mosdepth.nim(969) mosdepth
/home/brentp/src/mosdepth/mosdepth.nim(705) main
/nim-1.6.6/lib/system.nim(2621) []
/nim-1.6.6/lib/system/fatal.nim(53) sysFatal
Error: unhandled exception: value out of range: -149675 notin 0 .. 9223372036854775807 [RangeDefect]

(This is now on a different BAM file, but with the same issue. We accidentally deleted the other one...)

[brentp]

ok. thank you. I added some print info right at that location in the code. Can you run this one and show me what you see? I think you may have some very large coordinates in your bed file (larger than uint32 can hold)

mosdepth_debug2.gz

[Hoeze]

@brentp your linked binary is equal to the previous one, would you mind re-sending your binary?

[brentp]

mosdepth_debug2.gz
oops. here is is rebuilt and re-uploaded

[Hoeze]

These are the max start / end coordinates in the BED file:

# awk '{if($2>maxs) maxs=$2; if($3>maxe) maxe=$3} END{print maxs, maxe}' index0_germline1.converted_bed.bed
249210800 249212562

This is smaller than 4294967295, so should be fine, no?

[Hoeze]

@brentp Now it worked 👍
Here is the output:
mosdepth_debug2.log

[brentp]

ok. you have an interval that's off the end of the chromosome.

array-length: 248956423 r.start: 249106098 r.stop: 249106569

I'll add a check for that, but you can also remove that interval

[brentp]

here is a binary that skips intervals beyond end of chrom and prints a warning. would be great if you could give it a go

mosdepth_dev.gz

[Hoeze]

Awesome, your fix works!
Thanks a lot @brentp!

you have an interval that's off the end of the chromosome.

Not good. Probably a genome version mismatch...

I'll add a check for that, but you can also remove that interval

Would it be possible to make Mosdepth fail with the correct error message in that case?
I rather fix my input than miss a genome version mismatch...

[brentp]

let me think about that. I'd rather do a warning, but it could make sense to throw an error