On Mon, Dec 24, 2018 at 02:54:32PM +0100, Jeremie Courreges-Anglas wrote: > > As reported by Landry, > > < landry_> Error: > /usr/obj/ports/boost_1_66_0/fake-sparc64/usr/local/lib/libboost_locale-mt.a > does not exist > < landry_> Error: > /usr/obj/ports/boost_1_66_0/fake-sparc64/usr/local/lib/libboost_locale-mt.so.9.0 > does not exist > < landry_> fallout from otto's work ? > < landry_> - Boost.Locale needs either iconv or ICU library to be built. > < landry_> - iconv (libc) : no (cached) > < landry_> - iconv (separate) : no (cached) > < landry_> - icu : no (cached) > < landry_> - icu (lib64) : no (cached) > < landry_> faster bulks \o/ > < landry_> strange because the log still says - locale : > building > > (Same failure on macppc.) > > ICU detection for Boost.Locale fails and libboost_locale.so isn't built. > The breakage is due (at least) to char16_t now being used in the public > icu headers. This is not supported by base-gcc, and supported by > ports-gcc with -std=c++11 or newer. > > The diff below: > - restores the printing of compile commands, "-d+2", lost in the update > to boost-1.65.1. It's hard to debug anything without this. > - adds -std=gnu++11 to CXXFLAGS on archs using ports-gcc (which defaults > tp gnu++98) > - force the build to respect PORTS_PATH, so that g++ from ports-gcc is > used instead of g++ from base > > This fixes boost for me on sparc64, landry has put it earlier in sparc64 > and macppc bulks.
It built fine on sparc64, and hit an ICE on macppc. I've resumed this build, crossing fingers...
