Path: utzoo!telly!attcan!utgpu!jarvis.csri.toronto.edu!helios.physics.utoronto.ca!radio.astro.utoronto.ca!radio.astro!brian
From: brian@radio.astro.utoronto.ca (Brian Glendenning)
Newsgroups: gnu.gcc.bug
Subject: gcc -mfpa broken?
Message-ID: <BRIAN.89Dec14193848@radio.astro.utoronto.ca>
Date: 14 Dec 89 19:38:48 GMT
Sender: brian@radio.astro.utoronto.ca (Brian Glendenning)
Distribution: gnu
Organization: Radio Astronomy, Univeristy of Toronto
Lines: 38


Statistics: Sun 3 with floating point accelerator, SunOS 4.0.3,
gcc=1.36 (likely 1.35 also).

When I compile:

main ()
{
     float r;
     int i;

     i=10;
     r = 1.0*i;
}

I get the rather hideous error message:

 ./cc1 /usr/tmp/cca25486.cpp -quiet -dumpbase test.bad.c -mfpa -g -version -o /usr/tmp/cca25486.s
GNU C version 1.36 (68k, MIT syntax) compiled by GNU C version 1.36.
default target switches: -m68020 -mc68020 -m68881 -mbitfield
test.bad.c: In function main:
test.bad.c:8: The following insn was not recognizable:
(insn 13 11 14 (set (mem:SF (plus:SI (reg:SI 14)
               (const_int -4)))
       (float_truncate:SF (reg:DF 57))) -1 (nil)
   (nil))
./gcc: Program cc1 got fatal signal 6.

If I replace 1.0*i with just i it works. 1.35 also dies with signal
six, but no error message is emitted.

Is this some sort of installation error on my part or a real bug. If
it's a real bug is there a fix? Thanks!

Brian
--
	  Brian Glendenning - Radio astronomy, University of Toronto
brian@radio.astro.utoronto.ca uunet!utai!radio!brian  glendenn@utorphys.bitnet