Merge pull request #953 from benjjneb/taxonomy-segfault-fix

Taxonomy segfault fix

DESCRIPTION

@@ -8,8 +8,8 @@ Description: The dada2 package infers exact amplicon sequence variants (ASVs) fr
     removing substitution and chimera errors. Taxonomic classification is available
     via a native implementation of the RDP naive Bayesian classifier, and species-level
     assignment to 16S rRNA gene fragments by exact matching.
-Version: 1.14.0
-Date: 2019-04-22
+Version: 1.15.1
+Date: 2020-02-19
 Maintainer: Benjamin Callahan <[email protected]>
 Author: Benjamin Callahan <[email protected]>, Paul McMurdie, Susan Holmes
 License: LGPL-3

R/misc.R

@@ -331,7 +331,7 @@ bs1ham <- function(dd, ham=1) {
 
 #' @keywords internal
 getSRR <- Vectorize(function(run, outdir="sra", verbose=TRUE, ...) {
-  if(!grepl("^SRR[0-9]{6+}$", run)) stop("Requires SRA Run accessions in format: SRR1234567")
+  if(!grepl("^SRR[0-9]{6,}$", run)) stop("Requires SRA Run accessions in format: SRR1234567")
   if(!dir.exists(outdir)) dir.create(outdir)
   loc <- paste0("ftp-trace.ncbi.nlm.nih.gov/sra/sra-instant/reads/ByRun/sra/SRR/",
                 substr(run, 1, 6), "/", run, "/")

src/taxonomy.cpp

@@ -1,6 +1,8 @@
 #include "dada.h"
 #include <Rcpp.h>
 #include <RcppParallel.h>
+#include <random>
+
 using namespace Rcpp;
 
 // Gets kmer index
@@ -70,6 +72,9 @@ int get_best_genus(int *karray, double *out_logp, unsigned int arraylen, unsigne
   double p, logp, max_logp = 1.0; // Init value to be replaced on first iteration
   double rv; // Dummy random variable
   unsigned int nmax=0; // Number of times the current max logp has been seen
+  std::random_device rd;  //Will be used to obtain a seed for the random number engine
+  std::mt19937 gen(rd()); //Standard mersenne_twister_engine seeded with rd()
+  std::uniform_real_distribution<> cunif(0.0, 1.0);
 
   for(g=0;g<ngenus;g++) {
     genus_kv = &genus_kmers[g*n_kmers];
@@ -97,7 +102,8 @@ int get_best_genus(int *karray, double *out_logp, unsigned int arraylen, unsigne
       nmax=1;
     } else if (max_logp == logp) { // With uniform prob, store if equal to current max
       nmax++;
-      rv = (double) Rcpp::runif(1)[0];
+      rv = (double) cunif(gen);
+      ///!      rv = (double) Rcpp::runif(1)[0];
       if(rv < 1.0/nmax) {
         max_g = g;
       }