internal compiler error in gcc-2.95.1 on i686-pc-linux-gnu

[Jeffrey A Law]

  In message <000101bf1f25$0a72a600$c826d8c1@oemcomputer>you write:
  > When I compile the following preprocessed stub "test.i":
  > =======================================================================
  > 
  > # 1 "test.c"
  > typedef __complex__ double double_complex;
  > 
  > 
  > 
  > void
  > p (const double_complex *t, int n)
  > {
  >   int i;
  >   double s = ({ typeof ( t[n/2] ) arg = ( t[n/2] ); (__imag__ arg); }) ;
  >   for (i = 1; i < n/2; i++)
  >     s += 2* ({ typeof ( t[i] ) arg = ( t[i] ); (__imag__ arg); }) ;
  > }
  > 
  > =======================================================================
  > 
  > with the command "gcc -v -O -c test.i", I get the following message
  > 
  > =======================================================================
  > 
  > Reading specs from /usr/local/lib/gcc-lib/i686-pc-linux-gnu/2.95.1/specs
  > gcc version 2.95.1 19990816 (release)
  >  /usr/local/lib/gcc-lib/i686-pc-linux-gnu/2.95.1/cc1 test.i -quiet -O -vers
  > ion -o /tmp/ccBpbTRx.s
  > GNU C version 2.95.1 19990816 (release) (i686-pc-linux-gnu) compiled by GNU
  >  C version 2.95.1 19990816 (release).
  > test.c: In function `p':
  > test.c:12: Internal compiler error in `reg_overlap_mentioned_p', at rtlanal
  > .c:96
  > Please submit a full bug report.
  > See <URL: http://www.gnu.org/software/gcc/faq.html#bugreport > for instructio
  > ns.
Thanks.  This still fails in the development sources.  I haven't looked into
the root cause, but I have added this test to the regression testsuite.

Hopefully someone will look into the problem soon.
jeff