On 9 June 2014 00:16, Peter Michael Green <[email protected]> wrote: > > package: boost1.55 > version: 1.55.0+dfsg-1 > severity: important > x-debbugs-cc: [email protected] > > boost1.55 failed to build on arm64 with an internal compiler error. In an > attempt to fix this I disabled pch using the patch attatched in > boost1.55.debdiff (note: this patch disables it unconditionally, if applying > it in debian you would probablly want to make it architecture conditional). > > Disabling pch made the package compile without internal compiler errors. > Unfortunately it then failed with errors from dpkg-shlibdeps. >
That's good, let me devise a fix for that, throw it onto an Ubuntu porter box and if all good upload into Debian & Ubuntu. Thanks a lot for the idea. > Unfortunately I don't have a log of the failed build itself (and I'm not > rerunning the whole build, it took ages iwith qemu and I don't have access to > real hardware at this point) but I do have a log from rerunning debian/rules > binary after the failure (attatched as boost-binary.log) and it contained the > following errors. > > dpkg-shlibdeps: error: couldn't find library libboost_mpi_python.so.1.55.0 > needed by > debian/libboost-mpi-python1.55.0/usr/lib/python2.7/dist-packages/boost/mpi.aarch64-linux-gnu.so > (ELF format: 'elf64-littleaarch64'; RPATH: '') > dpkg-shlibdeps: error: couldn't find library libboost_python.so.1.55.0 needed > by > debian/libboost-mpi-python1.55.0/usr/lib/python2.7/dist-packages/boost/mpi.aarch64-linux-gnu.so > (ELF format: 'elf64-littleaarch64'; RPATH: '') > > Any thoughts on what the problem is? (I still have the build tree from the > failed build if anyone wants me to dig anything out of it) > The typo in the patch that disabled building mpi components. Another typo that doesn't disable pch during python/mpi builds. And missing set -e on our part to catch above. > I'm on holiday right now so if wookey or someone else wants to take over this > issue that would be great. Boost now seems to be one of the biggest blockers > for getting debian arm64 into a sane state. > > P.S. ubuntu split boost1.55 into two source packages (as part of their > main/universe split iirc), over there the main boost1.55 source package built > successfully on arm64 but the latest version of the boost-mpi-source1.55 > source package failed with an internal compiler error. I'll go file a bug > there when I get a bug number for this bug. > No need, I maintain boost in ubuntu, and was aware of the issue, but didn't have time to dig into which gcc upload regressed on this. Since there is no code changes in boost that would have caused boost arm64 FTBFS on ubuntu. > P.P.S. other versions of boost also FTBFS on debian arm64 but I have not > investigated why Regards, Dimitri. -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
