Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!ucsd!nmtsun!dwho
From: dwho@nmtsun.nmt.edu (David Olix)
Newsgroups: gnu.gcc.bug
Subject: Unrecognizable insn in gcc 1.36
Keywords: insn
Message-ID: <3854@nmtsun.nmt.edu>
Date: 27 Feb 90 22:37:42 GMT
Reply-To: dwho@jupiter.nmt.edu (David Olix)
Distribution: gnu
Organization: New Mexico Tech, Socorro NM
Lines: 46

The following piece of code crashes cc1 with fatal signal 6
when compiled and run as follows on a Sun 3/280 running
SunOs 4.0.3:

jupiter% gcc -v
gcc version 1.36

jupiter% cat test.c
int
main()
{
  int i;
  double foo;

  i = 1;
  foo = (double)i;
  return 0;
}

jupiter% gcc -mfpa test.c
test.c: In function main:
test.c:10: The following insn was not recognizable:
(insn 9 6 11 (set (mem:DF (plus:SI (reg:SI 14)
               (const_int -12)))
       (float:DF (mem:SI (plus:SI (reg:SI 14)
                   (const_int -4))))) -1 (nil)
   (nil))
gcc: Program cc1 got fatal signal 6.

Does this have something to do with the cast?  Replacing line 8 with:

  foo = (double)1;

doesn't result in the error.  Also, the error only occurs with the

  -mfpa

option on the command line.


Please e-mail to me on what you guys find on this since I have some non-trivial
code which relies on such a cast.  (And without the fpa installed, the program
runs _S_L_O_W_).

thank you
--David Olix