This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
HAVE_probe macro
- From: Larin Hennessy <larin at science dot oregonstate dot edu>
- To: gcc at gcc dot gnu dot org
- Date: Sun, 23 Feb 2003 12:03:35 -0800
- Subject: HAVE_probe macro
In the process of looking at undocumented macros, I ran across the
following in gcc/gcc/config/m68k/crds.h dating from 1997-02-15:
/* UNOS need stack probe :-( */
#if 0
#define HAVE_probe 1
#define gen_probe() gen_rtx_ASM_INPUT (VOIDmode, "tstb
-2048(sp)\t;probe\n")
#else
#undef NEED_PROBE
#define NEED_PROBE (-2048)
#endif
The only place that HAVE_probe seems to still be used is in
gcc/gcc/explow.c:
/* Some systems require a particular insn to refer to the stack
to make the pages exist. */
#ifdef HAVE_probe
if (HAVE_probe)
emit_insn (gen_probe ());
#endif
The only other refences I can find to HAVE_probe were from
FSFChangeLog.10 and 11 relating to removing HAVE_probe from
i386/winnt.h in 1995 and the change to m68k/crds.h in 1997.
In fact, the only references to NEED_PROBE and gen_probe seem to also
be in a few lines of files crds.h, plexus.h, m68k.c, m68k.h,
m68k.md, and explow.c. And none of these items appear to be documented.
m68k-crds-unos is still listed as a suported platform on the web pages,
but m68k-plexus-sysv is not currently listed.
I do not have of these boxes to test with, but the above code could not
have worked since 1997. Should the plexus code be removed,
and explow changed to use NEED_PROBE?
I would be happy to put together a patch, but I was wondering what the
correct course of action is in this case.
-Larin