Re: Machine-independent bug in cse, with analysis.

[Jeffrey A Law <law at cygnus dot com>]

  In message <199804180333.NAA04430@discus.anu.edu.au>you write:
  > 
  > Jeff said:
  > > Given time and appropriate incentive I could find bugs in every
  > > optimization pass in the compiler.
  > 
  > Here's one in CSE :-).  It's the last egcs bug that causes the glibc
  > test cases to fail on PowerPC.  It may be a known bug.
  > 
  > The following program, when compiled under (at least) egcs 1.0.2
  > configured for powerpc-unknown-linux-gnu, and under gcc 2.8.1
  > configured for sparc-sun-solaris2.6, with either gcc -O2 or
  > gcc -O1 -frerun-cse-after-loop, fails with exit code 1.  It should
  > (and does under plain -O1) succeed.  I would expect it appears under
  > all configurations, and I seem to remember it being in 2.7.2.1.
  > 
  > static int f(int) __attribute__((const));
  > int main()
  > {
  >    int f1, f2, x;
  >    x = 1; f1 = f(x);
  >    x = 2; f2 = f(x);
  >    return f1 == f2 ? 1 : 0;
  > }
  > static int f(int x) { return x; }
  > 
  > Any const definition for f will do.
This has been fixed :-)

jeff