Samtools "Eof Marker Is Absent" Error Only In Version 0.1.19 ?

Question

[Mike Axtell]

Using samtools version 0.1.18, the following bowtie alignment pipeline works well to create a sorted BAM file:

bowtie -v 0 -S genome.fasta test.fastq | samtools view -S -b -u - | samtools sort - test0.1.18
[samopen] SAM header is present: 2108 sequences.
# reads processed: 80094
# reads with at least one reported alignment: 70028 (87.43%)
# reads that failed to align: 10066 (12.57%)
Reported 70028 alignments to 1 output stream(s)

However, if I install and use samtools version 0.1.19, I get an error message claiming the input is truncated (2nd line below):

bowtie -v 0 -S genome.fasta test.fastq | samtools view -S -b -u - | samtools sort - test0.1.19
[bam_header_read] EOF marker is absent. The input is probably truncated.
[samopen] SAM header is present: 2108 sequences.
# reads processed: 80094
# reads with at least one reported alignment: 70028 (87.43%)
# reads that failed to align: 10066 (12.57%)
Reported 70028 alignments to 1 output stream(s)

I assume the error message is coming from samtools sort. But it is confusing, since both versions create identical results, including fully intact headers and properly formatted BAM alignments. Removing the -u switch in the samtools view call has no effect.

I'd like to fix the reporting of this error because a similar pipeline is used in some scripts of mine -- users who have samtools 0.1.19 are worried about the error message. I know I could just send the STDERR of the samtools sort to /dev/null, but I'd rather try and understand what is happening first ...

Answer 1

[Devon Ryan]

This is a known bug, you can ignore it. This was actually fixed at some point and then seems to have been accidentally reverted.

[Mike Axtell]

Thanks for digging that out for me - much appreciated

[Devon Ryan]

No problem. Given how frequently I use samtools (and its C API), I should probably check if the current devel branch still has this bug and submit a fix if not.

Answer 2

[scalabrin]

I got the same "warning" that appeared an error though! I digged into my data until I discovered that only the new version provides the "warning". Only successively I discovered this thread.

I hope it gets fixed soon and further people won't get affected by that. It is very annoying indeed!