[Samtools-help] Discrepancies in reported pileup read segments

[Lars B. <lar...@un...>]

Hi,

I've been running an analysis that depends on read segment boundaries as reported in pileup output from samtools mpileup. It appears that between samtools 1.2 (which was current when I first did the analyses), and samtools 1.3.1, the reported read segment starts and ends differ when run with "default" options (e.g. something like samtools mpileup -f ref.fa -d 100000 some.bam), in some cases dramatically.

As far as I've been able to figure it out, these seem to mostly be cases where a read segment starts with a mismatch, for instance a position where fields 4 and 5 would look like this with 1.2:
14	.............^~.
Now looks like this in 1.3.1:
15	.............^~.^~T

I've read through the release notes for 1.3 and 1.3.1 (https://github.com/samtools/samtools/releases), and while I see a few references to changes in mpileup there's nothing that obviously explains this difference in behavior (to me at any rate). I can see that somehow mismatches in read segments are being treated differently now, but am unclear on exactly how. 

Does anyone know exactly what has changed between these versions that would lead to this difference in behavior?

Thanks,
Lars