When built by the ld.gold linker, ntfs3g is incapable of mounting on my system. I'm afraid I couldn't coax much debugging information, but the following error is reported: "fuse: unknown option `'" Switching to ld.bfd and rebuilding ntfs3g resolved the issue. Reproducible: Always Steps to Reproduce: 1. binutils-config --linker ld.gold 2. emerge ntfs3g 3. ntfs-3g /dev/whatever /mnt/wherever I assume this is related to the existing problems fuse itself has with gold, but haven't debugged it further.
I can confirm this, found another report in the internez [1], can you add a warning like $(tc-getLD) --version | grep gold >/dev/null && ewarn "U R DOOMED" 'd appreciate it. [1] https://groups.google.com/forum/#!topic/tup-users/c53rfBsYCAk
Added a pkg_pretend that calls die when trying to build with gold (copied from grub's).
with gcc-4.8, there is -fuse-ld=bfd. we can add a toolchain-funcs.eclass helper to automate this which ebuilds would utilize.
(In reply to SpanKY from comment #3) > with gcc-4.8, there is -fuse-ld=bfd. we can add a toolchain-funcs.eclass > helper to automate this which ebuilds would utilize. Please do. It's really awful when ebuilds tell you to switch stuff like this.
(In reply to SpanKY from comment #3) > with gcc-4.8, there is -fuse-ld=bfd. we can add a toolchain-funcs.eclass > helper to automate this which ebuilds would utilize. If the build env respects the provided LD, I've been doing stuff like: src_compile() { emake LD="$(tc-getLD).bfd" } Which works on more than just gcc 4.8.
Can we apply some nicer workaround? Also could anyone report it to upstream? [1] [1] http://sourceforge.net/p/ntfs-3g/mailman/ntfs-3g-devel/
sys-fs/ntfs3g-2014.2.15-r1 now sets LD before running econf to make sure it uses the bfd linker rather than gold.