Hello Joseph, > On 21 Sep 2021, at 21:11, Joseph Myers <jos...@codesourcery.com> wrote: > > On Fri, 3 Sep 2021, Iain Sandoe wrote: > >> given that: >> >> a) this fixes Darwin x86-64 bootstrap which has been broken for more than 24h >> b) the patch is now Darwin-local. > > Actually, it's not Darwin-local. It uses __MACH__, which is also defined > for Hurd. And because sfp-machine.h gets included in files that aren't > specific to HFmode (and so aren't built with explicit -msse2), the build > for i686-gnu fails with: > > In file included from > /scratch/jmyers/glibc-bot/src/gcc/libgcc/config/i386/sfp-exceptions.c:25: > /scratch/jmyers/glibc-bot/src/gcc/libgcc/config/i386/sfp-machine.h:83:1: > error: unable to emulate 'HF' > 83 | typedef float alias_HFtype __attribute__ ((mode (HF))); > | ^~~~~~~ > > I think some conditional that is genuinely Darwin-specific should be used, > so that Hurd keeps using normal ELF aliases and doesn't get these HFmode > references in sfp-machine.h at all.
Sorry about this, (I usually use __APPLE__ as the flag, but the __MACH__ was already there in this case, I think). I’ll fix this by s/__MACH__/__APPLE__/ since in this case we definitely mean Mach-O rather than anything to do with the micro-kernel... (tomorrow now). Note, that I have a suspicion that there are (maybe a small number of) other places in the code where __MACH__ has been taken to mean mach-o is in use. thanks Iain
