{ stdenv , lib , fetchurl , perl , gmp , gf2x ? null # I asked the ntl maintainer weather or not to include gf2x by default: # > If I remember correctly, gf2x is now thread safe, so there's no reason not to use it. , withGf2x ? true , tune ? false # tune for current system; non reproducible and time consuming }: assert withGf2x -> gf2x != null; stdenv.mkDerivation rec { pname = "ntl"; version = "11.5.1"; src = fetchurl { url = "http://www.shoup.net/ntl/ntl-${version}.tar.gz"; sha256 = "sha256-IQ0GwxMGy8bq9oFEU8Vsd22djo3zbXTrMG9qUj0caoo="; }; buildInputs = [ gmp ]; nativeBuildInputs = [ perl # needed for ./configure ]; sourceRoot = "${pname}-${version}/src"; enableParallelBuilding = true; dontAddPrefix = true; # DEF_PREFIX instead # reference: http://shoup.net/ntl/doc/tour-unix.html configureFlags = [ "DEF_PREFIX=$(out)" "SHARED=on" # genereate a shared library (as well as static) "NATIVE=off" # don't target code to current hardware (reproducibility, portability) "TUNE=${ if tune then "auto" else if stdenv.targetPlatform.isx86 then "x86" # "chooses options that should be well suited for most x86 platforms" else "generic" # "chooses options that should be OK for most platforms" }" "CXX=${stdenv.cc.targetPrefix}c++" ] ++ lib.optionals withGf2x [ "NTL_GF2X_LIB=on" "GF2X_PREFIX=${gf2x}" ]; doCheck = true; # takes some time meta = with lib; { description = "A Library for doing Number Theory"; longDescription = '' NTL is a high-performance, portable C++ library providing data structures and algorithms for manipulating signed, arbitrary length integers, and for vectors, matrices, and polynomials over the integers and over finite fields. ''; # Upstream contact: maintainer is victorshoup on GitHub. Alternatively the # email listed on the homepage. homepage = "http://www.shoup.net/ntl/"; # also locally at "${src}/doc/tour-changes.html"; changelog = "https://www.shoup.net/ntl/doc/tour-changes.html"; maintainers = teams.sage.members; license = licenses.gpl2Plus; platforms = platforms.all; }; }