On 8 December 2014 at 18:38, terrance savitsky wrote: | Hello Rcpp Developers, I had earlier forwarded a report that initialization of | arma::SpMat objects through as<arma::spmat>() in my package generated a memory | management fault in valgrind for my package, growfunctions. The issue was | sourced (not by me) and a fix was incoporated at RcppArmadillo version 0.4.400 | (and I am now using 0.4.550.1.0 in this latest build). I will remove use of | the SpMat type from my package as it is not extensively employed, but perhaps | the issue warrants further investigation for RcppArmadillo.
One day we may come across someone who uses sparse matrices and also has both the skill and motivation to investigate this. As I don't use them, I am not likely to find or make the time for it. Thanks for he report, Dirk | | Thank you, Terrance Savitsky | ---------- Forwarded message ---------- | From: Prof Brian Ripley <rip...@stats.ox.ac.uk> | Date: Mon, Dec 8, 2014 at 5:33 PM | Subject: Re: CRAN submission growfunctions 0.12 | To: Terrance Savitsky <tds...@gmail.com>, CRAN <c...@r-project.org> | | | On 08/12/2014 22:04, Prof Brian Ripley wrote: | | | | And valgrind is still reporting issues in Armadillo, e.g. | | ==30106== Conditional jump or move depends on uninitialised value(s)==30106== | at 0x15EE8D9A: operator!= (test-3.2/RcppArmadillo/include/armadillo_bits/ | SpMat_iterators_meat.hpp:239) | ==30106== by 0x15EE8D9A: arma::enable_if2<(arma::is_arma_type<arma::Gen | <arma::Mat<double>, arma::gen_ones_diag> >::value&&arma::is_arma_sparse_type | <arma::SpM | at<double> >::value)&&arma::is_same_type<arma::Gen<arma::Mat<double>, | arma::gen_ones_diag>::elem_type, arma::SpMat<double>::elem_type>::value, | arma::Mat<arma::Gen<arma::Mat<double>, arma::gen_ones_diag>::elem_type> | >::result arma::operator- | <arma::Gen<arma::Mat<double>, arma::gen_ones_diag>, arma::SpMat<double> > | (arma::Gen<arma::Mat<double>, arma::gen_ones_diag> const&, arma::SpMat<double> | const&) | (test-3.2/RcppArmadillo/include/armadillo_bits/operator_minus.hpp:281) | ==30106== by 0x15EE0FB1: IGMRFDPMIX (packages/incoming/growfunctions.Rcheck/ | 00_pkg_src/growfunctions/src/dpmix.cpp:66) | ==30106== by 0x480C3C: R_doDotCall (svn/R-devel/src/main/dotcode.c:655) | ==30106== by 0x48158F: do_dotcall (svn/R-devel/src/main/dotcode.c:1251) | | /main/eval.c:2034)==30106== Uninitialised value was created by a heap | allocation | ==30106== at 0x4A09190: memalign (/data/gannet/ripley/Sources/srcF20/ | valgrind-3.10.1/coregrind/m_replacemalloc/vg_replace_malloc.c:760) | ==30106== by 0x4A092A7: posix_memalign (/data/gannet/ripley/Sources/srcF20/ | valgrind-3.10.1/coregrind/m_replacemalloc/vg_replace_malloc.c:913) | ==30106== by 0x15EECC85: acquire<unsigned int> (test-3.2/RcppArmadillo/ | includ | e/armadillo_bits/memory.hpp:75) | ==30106== by 0x15EECC85: acquire_chunked<unsigned int> (test-3.2/ | RcppArmadill | o/include/armadillo_bits/memory.hpp:111) | ==30106== by 0x15EECC85: Rcpp::traits::Exporter<arma::SpMat<double> >::get() | (test-3.2/RcppArmadillo/include/RcppArmadilloAs.h:101) | ==30106== by 0x15EE1096: as<arma::SpMat<double> > (test-3.2/Rcpp/include/ | Rcpp | /as.h:81) | ==30106== by 0x15EE1096: as<arma::SpMat<double> > (test-3.2/Rcpp/include/ | Rcpp/as.h:144) | ==30106== by 0x15EE1096: IGMRFDPMIX (packages/incoming/growfunctions.Rcheck/ | 00_pkg_src/growfunctions/src/dpmix.cpp:63) | ==30106== by 0x480C3C: R_doDotCall (svn/R-devel/src/main/dotcode.c:655) | ==30106== by 0x48158F: do_dotcall (svn/R-devel/src/main/dotcode.c:1251) | | | | -- | Brian D. Ripley, rip...@stats.ox.ac.uk | Emeritus Professor of Applied Statistics, University of Oxford | 1 South Parks Road, Oxford OX1 3TG, UK | | | | -- | Thank you, Terrance Savitsky | _______________________________________________ | Rcpp-devel mailing list | Rcpp-devel@lists.r-forge.r-project.org | https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel -- http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org _______________________________________________ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
