> I poked at this a bit more. I can reproduce the problem by using > -mfpmath=387 on dromedary's host (fairly old 32-bit macOS); although > I also get half a dozen *other* failures in the core regression tests, > mostly around detection of float overflow. So I'm not quite sure that > this is comparable. But at any rate, I tracked the core of the problem > to pg_hypot:
I couldn't test if it helps, but another solution may be is to rip out pg_hypot() in favour of the libc implementation. This was discussed in detail as part of "Improve geometric types" thread. Last comment about it: https://www.postgresql.org/message-id/9223.1507039405%40sss.pgh.pa.us