segmentation fault bug

[Jeffrey A Law]

  In message < 36338F72.5C3B41FC@inka.de >you write:
  > Hi,
  > 
  > below is a description of a bug which might be a bug in egcs.
  > 
  > 1. I have the following source ("testn.cc"):
  > 
  >      #include <stdio.h>
  >      #include <math.h>
  >   
  >      main()
  >      {
  >          float v = fabs(-1.); // A
  >          printf("a");         // B
  >          pow(1., 5.);         // C
  >      }
  > 
  > 2. I compile it with "g++ -O testn.cc".
  > 3. When I call "a.out" I get the following error message 
  >    (this should not happen and therefore seems to be a bug):
  >    "Segmentation fault"
  > 4. Hardware: AMD K6-2 300 or P90
  > 5. OS: Debian LINUX 2.0 with a 2.0.34 kernel
  > 6. g++ -v:
  >      Reading specs from /usr/lib/gcc-lib/i486-linux/egcs-2.90.29/specs
  >      gcc version egcs-2.90.29 980515 (egcs-1.0.3 release)
  > 7. The bug does *not* occur:
  >      - when I remove either A, B, or C.
  >      - with gcc version 2.7.2.3.
  >      - when I copmpile "testn.cc" without the "-O" option.
  > 
  > Do you have any clues if this bug might be my fault, a egcs bug,
  > or some kind of LINUX bug?
It is most likely an egcs bug.  The pow function in glibc2 tickles a bug in
egcs-1.0.x compilers.  egcs-1.1 should do the right thing.

jeff