On 2026/07/24 14:51, Stuart Henderson wrote: > On 2026/07/23 22:07, Theo Buehler wrote: > > I think fixing it this way means that we need to add > > > > USE_NOBTCFI-aarch64 = Yes# fortran > > > > to quite a few ports, unless someone can confirm that this is no longer > > needed with lang/gcc/15 - as the XXX in fortran.port.mk suggests. > > I've just run 'make test' with gfortran 15 on Apple m2 on a few ports > which currently have "USE_NOBTCFI-aarch64 = Yes# fortran" with the > USE_NOBTCFI removed, and not seeing any problems: > > astro/wcslib: 1 unrelated failure in units(1)-related test > geo/eccodes: clean > graphics/gegl04: clean > astro/gnuastro: some failures because programs-builts/.libs/asttable, > ../astfits, ../astmkcatalog were not built but seem > unrelated > > (the markers in py-* ones are pointless anyway I think, because it's > the python binary itself that counts, and that's already marked > USE_NOBTCFI ...) > > So I think perhaps we can get rid of this now. (Since the main > _SYSTEM_VERSION was just bumped anyway, bumping _SYSTEM_VERSION-aarch64 > for this change would probably make sense as it probably won't hurt > users, and easier than figuring out all the bumps). But for an > additional test I want to try gimp at runtime. I'll leave it building > while I look for either the jetkvm or a way to connect the machine up > to a monitor whichever happens first..
Tried gimp-3.2 now, that's working ok. I think it would make sense to drop these USE_NOBTCFI-aarch64.
