long double trouble with gcc 4.0
I have recently upgraded to Xcode 2.2.1 from 2.0. Suddenly, reliable C code (a simulation model), that uses lots of long doubles, has started producing unusual output.
The model tracks frequencies stored in arrays of long doubles. I first noticed that frequecies that should be 1 are 2, which should be impossible. On closer inspection, many of the other numbers are not what I would expect, while some are what I would expect.
I have determined that this is a gcc 4.0 specific problem. Compiling with 3.3 or 2.95 (on various G4 and G5 PPC machines) gets rid of the problem.
It may be just the output that is messed up, as part of the simulation still reach its stopping point (based on the frequencies) in the same number of generations. Also, when I check values in the debugger, they seem fine. I am printing with printf and fprintf statements using %Le format .
Any ideas what is going on?
The model tracks frequencies stored in arrays of long doubles. I first noticed that frequecies that should be 1 are 2, which should be impossible. On closer inspection, many of the other numbers are not what I would expect, while some are what I would expect.
I have determined that this is a gcc 4.0 specific problem. Compiling with 3.3 or 2.95 (on various G4 and G5 PPC machines) gets rid of the problem.
It may be just the output that is messed up, as part of the simulation still reach its stopping point (based on the frequencies) in the same number of generations. Also, when I check values in the debugger, they seem fine. I am printing with printf and fprintf statements using %Le format .
Any ideas what is going on?
iMacG5 and PowerMacG5 PPC machines, Mac OS X (10.4.6), Xcode 2.2.1