Anton Korobeynikov

Well, this looks like an NFS issue as at the time of error all files are supposed to be already written. Its not like "SPAdes checks the presence early" –...

> Could the overflown number show the file actually existed but it's presence is mis-interpreted? It's just -1 which is used to signal an error

Well, the temp dir is transient and the output dir is used to store important intermediate files. Per NFS specification, the sync is done at the file close (https://linux.die.net/man/5/nfs), so...

SPAdes supports arbitrary number of input libraries. See https://cab.spbu.ru/files/release3.15.0/manual.html#sec3.1 (section about YAML dataset specification) for more results.

> I recompiled from sources and now I could at least start loading-in the FASTQ data again (was getting OS error -11 as others already reported). Let's see, if the...

But this is a *single* library consisting many files. Not multiple libraries – quite a huge difference

Well, your input contains almost 30 billion distinct k-mers. I think this is one of the largest datasets I saw to-date. BayesHammer would require up to 40-50 bytes per k-mer,...

Probably related to system pyaml used in Debian? SPAdes ships (supports and uses) with bundled pyyaml. No outside pyaml is supported.

This patch is actually the perfect example of why we do this. We do not have bandwidth of supporting multiple versions of the dependencies scattered over multiple OSes, distributions, etc....

There is no such thing as "loader" there at all. It's a breaking change from pyaml 5.1. So you'd need to introduce the version check and / or some kind...