Re: RFA: USE_STRUCT_CONVENTION for NetBSD/i386

[Mark Kettenis <kettenis at wins dot uva dot nl>]

   Date: Fri, 24 Mar 2000 13:57:20 +1100
   From: Andrew Cagney <ac131313@cygnus.com>

   "J.T. Conklin" wrote:

   > ! int
   > ! i386nbsd_use_struct_convention (int gcc_p, struct type *type)
   > ! {
   > !   return !(TYPE_LENGTH (type) == 1
   > !          || TYPE_LENGTH (type) == 2
   > !          || TYPE_LENGTH (type) == 4
   > !          || TYPE_LENGTH (type) == 8);
   > ! }
   > !

   I'd add a FIXME pointing out known problems with the function. 
   Otherwize ok.

Well, it's not the i386nbsd_use_struct_convention that's the problem,
so a FIXME there would not be very useful.

   I assume Mark is trying to figure out why NetBSD can't simply use a
   common i386_use_struct_convention function.  I take it from your answer,
   that NetBSD's convention isn't 100% standard (where standard would mean
   what someone other than Linux or *BSD did.  Solaris/x86?).

Basically, what we're talking about is the difference between the
System V ABI (Linux/ELF, Solaris/x86 and other System V-derived
systems) and the old, more efficient GCC convention used by NetBSD and
FreeBSD and Linux/a.out.  For the System V ABI,
generic_use_struct_convention does the right thing.  But we need a
different one for the other systems.

I'm a bit surprised by the fact that small struct's containing one
float are returned in the FP registers, so I want to investigate that
a bit, but I think it should be possible to have
i386_extract_return_value do the right thing there too.

Since this didn't work in the past anyway, I don't think it is release
critical though.

Mark