Closing .fsa and .fna files

[almahmoud]

This problem came about when running prokka in Galaxy on Kubernetes with outdir on an NFS volume.
prokka would only work the second time it was called, the first time failing with

[barrnap] ERROR: bad line in nhmmer output - Fatal exception (source file esl_buffer.c, line 1599):
java.lang.NullPointerException
	at FASTAReader.isFASTA(FASTAReader.java:32)
	at CRISPRFinder.goCRISPRFinder(CRISPRFinder.java:114)
	at minced.main(minced.java:259)

I then noticed that deleting outdir after the first run made the second run also fail, then noticed that specifically having outdir/prokka.fsa and outdir/prokka.fna alone was enough to make it pass. I believe this is because on the NFS, the files will not appear when writing without closing, but on a local filesystem it isn't a problem. After adding the following two lines, closing the files, prokka passed on the NFS!

[almahmoud]

I see the tests are failing, but I don't think that it is related to the PR. Anything I should do about it or is it okay because it's a relatively small change?

[tseemann]

Ideally you would not use NFS for your actual computational folder. I would recommend using --outdir as a file in your fast scratch storage (eg. /tmp ?) and then copying the results to NFS afterwards.

The files are automatically closed when the script ends. That patch occurs at the end of prokka where it writes out all the output files. I am not sure how what would carry over? It sounds like your NFS might be mounted in a special (inconsistent) manner? async perhaps?