"Segmentation fault (core dumped)" using sickle for trimming, depending on options used

Question

[rioualen]

Hello,

I'm running into "Segmentation fault (core dumped)" error when running a workflow with (or without) snakemake. I've used the same data, same tools and same parameters on several machines without a problem before.

Now the problem arises on a Ubuntu 14.04 virtual machine (4 CPU, 8Go RAM), when executing sickle. I've noticed that it depends on the parameters that I use, even when I'm specifying parameters that are supposed to be the default ones.

root@vm0083:~/mydisk/ChIP-seq_SE_GSE20870# sickle se -f fastq/GSM521934/GSM521934.fastq -t sanger -o sickle.fastq -q 20 -l 20

FastQ records kept: 2449731
FastQ records discarded: 1137314

root@vm0083:~/mydisk/ChIP-seq_SE_GSE20870# sickle se -f fastq/GSM521934/GSM521934.fastq -t sanger -o sickle.fastq -q 20 --length-threshold 20
Segmentation fault (core dumped)

root@vm0083:~/mydisk/ChIP-seq_SE_GSE20870# sickle se -f fastq/GSM521934/GSM521934.fastq -t sanger -o sickle.fastq --qual-threshold 20 -l 20
Segmentation fault (core dumped)

Parameters names and default values:

root@vm0083:~/mydisk/ChIP-seq_SE_GSE20870# sickle se -h
sickle: invalid option -- 'h'

Usage: sickle se [options] -f <fastq sequence file> -t <quality type> -o <trimmed fastq file>

Options:
-f, --fastq-file, Input fastq file (required)
-t, --qual-type, Type of quality values (solexa (CASAVA < 1.3), illumina (CASAVA 1.3 to 1.7), sanger (which is CASAVA >= 1.8)) (required)
-o, --output-file, Output trimmed fastq file (required)
-q, --qual-threshold, Threshold for trimming based on average quality in a window. Default 20.
-l, --length-threshold, Threshold to keep a read based on length after trimming. Default 20.
-x, --no-fiveprime, Don't do five prime trimming.
-n, --trunc-n, Truncate sequences at position of first N.
-g, --gzip-output, Output gzipped files.
--quiet, Don't print out any trimming information
--help, display this help and exit
--version, output version information and exit

I would like to know if this issue is likely related to sickle itself, or the VM I'm using, or something else...

Thank you

[kloetzl]

Verify that you input files are in the correct format. A lot of tools don't do proper parsing.

[rioualen]

They look normal, and again I've ran this tool on these specific files lots of times before.

[Matteo Schiavinato]

My guess is that the virtual machine has much less CPU and RAM allowed to be used, therefore yielding the segmentation fault.

[rioualen]

The VM has 4 CPUs and 8Go of RAM, and I've ran it successfully in smaller VMs before. Also, I don't get why just changing the name of the parameter is generating this type of error.

[Matteo Schiavinato]

Could you paste the code block where these arguments are parsed? (if python argparse, perhaps paste also the head of the script where they are declared!)

[rioualen]

int single_qual_threshold = 20;
int single_length_threshold = 20;

static struct option single_long_options[] = {
    {"fastq-file", required_argument, 0, 'f'},
    {"output-file", required_argument, 0, 'o'},
    {"qual-type", required_argument, 0, 't'},
    {"qual-threshold", required_argument, 0, 'q'},
    {"length-threshold", required_argument, 0, 'l'},
    {"no-fiveprime", no_argument, 0, 'x'},
    {"discard-n", no_argument, 0, 'n'},
    {"gzip-output", no_argument, 0, 'g'},
    {"quiet", no_argument, 0, 'z'},
    {GETOPT_HELP_OPTION_DECL},
    {GETOPT_VERSION_OPTION_DECL},
    {NULL, 0, NULL, 0}
};

[rioualen]

    case 'q':
        single_qual_threshold = atoi(optarg);
        if (single_qual_threshold < 0) {
            fprintf(stderr, "Quality threshold must be >= 0\n");
            return EXIT_FAILURE;
        }
        break;

    case 'l':
        single_length_threshold = atoi(optarg);
        if (single_length_threshold < 0) {
            fprintf(stderr, "Length threshold must be >= 0\n");
            return EXIT_FAILURE;
        }
        break;

[Matteo Schiavinato]

Are you sure this is the code block? This seems to me the code section where they check if quality and length are a positive number before doing analysis.

It should be something like:

if (single_qual_threshold >= qual_threshold) {
 > PROCEED <
} else {
 > DISCARD <
}

[rioualen]

I'm sorry but I'm not the one who wrote Sickle, that's why I actually need help...

[Matteo Schiavinato]

Programs are made on test data. Actual data are different, and not always working in programs that should be made for every kind of data. I modify scripts all the time to adapt them to my needs! Hope this is not the case, but sometimes you just happen to spot a bug.

Answer 1

[rioualen]

In case someone else is running into this issue, using shortcuts for params names avoids the problem: -q instead of --qual-threshold and -l instead of --length-threshold, specifically.

So in my case the problem was not related to my VM, but to the fact that my VM was using another version of my code which included full params names.