adapter detection gets stuck and uses all memory on ONT data

[mcolpus]

Hello,

I'm using fastp in a pipeline and found that some samples were giving out of memory errors when going through fastp. I've attached a small example: small.fq.gz

I'm using 0.24.1 and I just run:

fastp -i small.fq.gz  -o clean.fastq.gz

It displays:

Detecting adapter sequence for read1...

but gets stuck whilst increasingly using more ram (100G+)

There are some very short reads so I tried filtering any <100bp long first, but that doesn't fix the issue.

If I disable adapter trimming then the sample does run completely fine.

Thanks, Let me know if you need any more info

[ckrushton]

I've been running into some memory issues as well (albeit on short read sequencing data) and took a quick look at the adapter trimming logic using your test file. I encountered the same spike in memory usage, and it looks like its occurring in the Evaluator::getAdapterWithSeed function, where some FASTQ entries are having dozens or hundreds of candidate adapter hits, which are all being stored and used to determine the actual adapter sequence.

[sfchen]

I will fix this soon