On Sat, Dec 10, 2022 at 3:50 PM Tom Lane <t...@sss.pgh.pa.us> wrote: > Recently, buildfarm member elver has started spewing literally > thousands of $SUBJECT: > > elver | 2022-12-10 01:17:29 | > ../../src/include/utils/float.h:223:33: warning: due to lvalue conversion of > the controlling expression, association of type 'volatile float' will never > be selected because it is qualified [-Wunreachable-code-generic-assoc] > elver | 2022-12-10 01:17:29 | > ../../src/include/utils/float.h:223:33: warning: due to lvalue conversion of > the controlling expression, association of type 'volatile double' will never > be selected because it is qualified [-Wunreachable-code-generic-assoc] > elver | 2022-12-10 01:17:29 | > ../../src/include/utils/float.h:223:33: warning: due to lvalue conversion of > the controlling expression, association of type 'volatile long double' will > never be selected because it is qualified [-Wunreachable-code-generic-assoc] > [ etc etc, about 9200 times per build ] > > I have no idea what that means, and consulting the clang documentation > didn't leave me much wiser. I do see that a lot of these seem to be > associated with isnan() calls, which makes me guess that clang does > not play nice with however isnan() is declared on that box.
It was using LLVM and clang 15 for the JIT support (the base compiler cc is clang 13 on this system, but CLANG is set to 15 for the .bc files, to match the LLVM version). Apparently clang 15 started issuing a new warning for math.h. That header has since been adjusted[1] to fix that, but that's not going to show up in the release that elver's using for a while. I've told it to use LLVM/clang 14 instead for now; let's see if that helps. [1] https://github.com/freebsd/freebsd-src/commit/8432a5a4fa3c4f34acf6136a9077b9ab7bbd723e
