[RFA 1/3] Initial support for variant parts

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tom@tromey.com> writes:

Tom> It seems to me that using a wide-enough scalar here would be enough for
Tom> all the cases you list, provided that sign extension was handled
Tom> properly.  I'm not 100% sure this does handle sign extension though -- I
Tom> can update the test for that.

I did this, and I think my patch is ok here -- but this revealed a
latent bug in unpack_bits_as_long.  That function was not doing sign
extension in the case where the field was not a bitfield.  This seems
incorrect to me, so I will add another patch to my series to fix this.

Tom