>    (though if support for them is ever put in here, they should be,
>    but for i486-en only...
> 
> I don't follow your logic there.  My 386 certainly thinks it knows about
> FS and GS, and theoretically we should be making them as accessible as
> any other registers.

my logic was faulty here.  i know fortunately little about the i386
architecture, and (for some reason) i'd thought that fs and gs were
i486-specific.

>  Now it's true that we don't really use them for any
> `normal' applications, but that's not an adequate answer; who the Hell
> knows what, e.g., WINE programs might do, and there's no documentation
> that says their use is forbidden.

no documentation says that their use is forbidden, but the i386 header
files had O(3) different sets of "registers which are available to the
user" via ptrace() and whatnot.  since the ptrace code uses this
struct, and doesn't *ever* set the values, i don't see them as being
'safe' to have as part of the struct.  Since i was cleaning that cruft
up a bit anyway, they went away.


cgd

------------------------------------------------------------------------------