infinite loop in kseq_read for gz files that fail crc check

[pmelsted]

When a gz file fails the crc check kseq_read enters an infinite loop.

This happens in https://github.com/attractivechaos/klib/blob/master/kseq.h#L101

The __read method when using gzread returns -1 when the crc check fails and the for loop never exits.

I have reproduced this bug using the test program from your website and the two supplied files. They are identical except I messed with a CRC byte in t.fastq.gz

reads_2.fastq.gz
t.fastq.gz

Interestingly seqtk does not fail, although it doesn't read all the lines

> seqtk seq t.fastq.gz | wc -l
   39842
> seqtk seq reads_2.fastq.gz| wc -l
   40000

and gzcat (mac version of zcat) notices the error, but it returns the original data.

> gzcat reads_2.fastq.gz > /dev/null
> gzcat t.fastq.gz>  /dev/null
gzcat: invalid compressed data--crc error
gzcat: t.fastq.gz: uncompress failed

[pmelsted]

Another question is what is the correct behaviour of kseq in cases where __read returns a -1. Should it return a -1: end-of-file or another error code, say -3: file-error?

[lh3]

Yes, return value from read/gzread should be checked. Could you send a pull request? Thanks a lot for this!