I'm not terribly worried about this, but in case it's something that upstream would want to backport into the gcc-14.x branch: The test suite for sci-libs/qd-2.3.24 fails on riscv, but only when compiled with gcc-14 and optimization enabled (-O2 or higher). Lower optimization levels are OK, and so are gcc-15 and clang-19 at -O3. So whatever the problem is, it was likely fixed in gcc-15 already. FWIW my gcc-15 is currently Gentoo Hardened 15.0.0_pre20241208-r1 p31). It's fairly easy to reproduce outside of portage as well: $ ./configure --disable-static --enable-ieee-add --disable-sloppy-mul --disable-sloppy-div --enable-inline --disable-fortran $ make check The one test failure that arises is for the exponential function (exp) within qd_test: Test 7. (Sanity check for exp). Test FAILED.
@riscv people: I'm happy to give advice on how to look at it and reduce it and so on but I'm unlikely to run it down myself.
Time has been flying and ~riscv is unstable, so I'm content to just wait for gcc-15 in ~arch. No one is clamoring for this to be keyworded -- it was only on my list as an optional dependency of a dependency of SageMath.
Yeah, normally, I'd want to run it down for my own peace of mind out of correctness, but riscv moves soooo quickly upstream too that I wouldn't be surprised if it ended up being fixed by some large change.
