Tested with the ppc64el cross compiler (gcc 15.3) + qemu-user in a stonking chroot, 4ti2 1.6.15+ds-1, test/ppi:
- -O3 -flto: FAILED - -O2 (no LTO): FAILED - -O0: FAILED - -O2 -fsigned-char: FAILED - native amd64 -O2: PASSED So it's not the optimisation flags (nor char signedness): 4ti2-ppi deterministically computes a different, larger result on ppc64el (e.g. ppi5: 50 vectors instead of 47; ppi10: 3404 vs 2466; ppi16: 190377 vs 99329), with extra non-reduced vectors such as '-1 -1 0 0 1'. Since Debian's ppc64el build passes, the remaining Ubuntu-specific differences are the toolchain/runtime (IEEE 128-bit long double ABI on ppc64el, gcc 15.3, glibc 2.44, gmp). Not investigated further yet. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2170794 Title: 4ti2 FTBFS on ppc64el: 4ti2-ppi test fails (likely -O3/LTO), blocks barvinok To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/4ti2/+bug/2170794/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
