libgcc soft float functions failure on aarch64

[nia <nia%NetBSD.org@localhost>]

Occasionally I see errors like this in pkgsrc when compiling for
aarch64:

- undefined reference to `__fixdfti'
- undefined reference to `__fixunsdfti'

Apparently these are GCC soft float library routines for floating point
emulation, but (a) this is aarch64, we have floating point (b) no
compiler flags indicating we want soft float are being used.

These are both functions that take a double as an argument and return
a 64-bit integer.

Can anyone help me understand what's happening?