This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Updated PATCH for expr.c problems (MIPS, 386), 1.1.2 andcurrent
- To: law at cygnus dot com
- Subject: Re: Updated PATCH for expr.c problems (MIPS, 386), 1.1.2 andcurrent
- From: Hans-Peter Nilsson <hp at bitrange dot com>
- Date: Mon, 12 Apr 1999 06:27:49 -0400 (EDT)
- cc: plai at Lynx dot COM, egcs-patches at egcs dot cygnus dot com
On Mon, 12 Apr 1999, Jeffrey A Law wrote:
> In message <Pine.BSF.4.02A.9904120519480.27790-100000@dair.pair.com>you write
> > Not if your point is that the mips address recognition is
> > responsible.
> Yes. That is still my point. And apparently you still do not understand.
I understand that force_reg gets a (plus reg const_int) on multiple
targets, and that a (plus ...) is invalid input to force_reg.
I don't understand why you don't agree that that is really a
target-independent bug.
> > Incorrect, I get it on i686-pc-linux-gnulibc1 *native* too, just
> > as I said and showed in
> > <URL:http://egcs.cygnus.com/ml/egcs-patches/1999-03/msg00462.html>.
> > The same thing happens with current CVS.
> No. I just tried it on the x86. No problems at all.
I did too and guess what... Still same problem.
No *visible* problems though; no crash, no incorrect assembly code.
As I said, force_reg (and emit_move_insn) receives a (plus ...)
where it should have got a reg, mem or const; that's what can be
seen on other targets (than the mipsel-unknown-netbsd).
If you were looking for a crash, you would not have seen it.
If you were doing a "break emit_move_insn if y->code == PLUS" you should
have seen it.
brgds, H-P