--enable-vtable-verify causes ICEs on ppc64, so disable that,
also add arch specific bits everywhere else
the new patch is a backport from current gcc9/8.x tree, fixing
errors about "__float128 is not supported on this target" when
c++ math library on ppc64 and possibly other platforms (manifests
for example when building the bullet physics library)
we also need a patch for ppc64 musl, otherwise it will fail in
decimal float stuff, that one is taken from alpine linux
also serial-configure is broken for some cross targets, so
disable it
[ci skip]
9e4b02efa9