[Bug target/11251] [3.3/3.4 Regression] g77 complex number program produces wrong code when optimising

[jfb at npl dot co dot uk]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11251



------- Additional Comments From jfb at npl dot co dot uk  2003-06-23 10:08 -------
Yes, it must be a bug, round off errors could not cause this.

Also, adding a 'write' statement at the end of subroutine 'cross' will cause 
the program to work correctly! I should also point out that this problem only 
occurs with complex, not real numbers a,b,c. Hence it does not occur in C/C++.

Also the fault only occurs when you do something like "call cross (c,c,b)" ie 
passing the same variable to two dummy variables in the subroutine causes the 
failure. Therefore the problem seems to be concerned with linking the subs 
together.