This is the mail archive of the
gdb-patches@sourceware.cygnus.com
mailing list for the GDB project.
Re: [PATCH]: procfs fix for 64-bits (ATTN: Nick Duffek
- To: "Nick Duffek <Nick Duffek" <nsd at cygnus dot com>
- Subject: Re: [PATCH]: procfs fix for 64-bits (ATTN: Nick Duffek
- From: Michael Snyder <msnyder at cygnus dot com>
- Date: Wed, 03 May 2000 14:13:53 -0700
- CC: gdb-patches at sourceware dot cygnus dot com
- Organization: Cygnus Solutions
- References: <200005031955.MAA14413@seadog.cygnus.com> <200005032107.VAA06487@nog.bosbc.com>
Nick Duffek
>
> On 3-May-2000, msnyder@cygnus.com wrote:
>
> >Solaris declares this set of /proc parameters to be long.
> >UnixWare declares them to be int. For Sparc32 it doesn't
> >make a difference
>
> Likewise for current versions of UnixWare, on which sizeof(int) ==
> sizeof(long).
>
> However, the change would make procfs.c noncomformant with the UnixWare
> proc interface, which -- according to the proc(4) man page --requires int
> control messages. This may become a problem if UnixWare becomes 64-bit.
>
> >> If this is necessary, we should try to devise an autoconf test.
> >
> >Yep, which is exactly why I would avoid it if I can.
>
> What about using "CMDTYPE" instead of "long", and provisionally #defining
> CMDTYPE in procfs.c as long?
I'd rather use a typedef, but yes. Sounds like we might
be better off going that way. It'll have to be an autoconf
test. Moreover, if it winds up being different between
UW-32 and UW-64, and if we want a UW-64 debugger to be
able to debug a UW-32 app, we may wind up having to
gdbarch it, or define a whole separate set of interfaces.
For now, either a long or an autoconf typedef will probably
be fine.