Re: SCO compiler, cc or rcc? rounding errors?

In article <679075402wnr@bash.demon.co.uk>, bashir Akram <bash@bash.demon.co.uk> wrote: >Machine A is the development machine, running Sco on a Dec. >Machine B is the "Live" machine. running Sco on a compaq. > >On machine A, c4gl uses cc to compile programs. >On machine B, c4gl is using rcc to compile programs. Sounds like they are running different versions. The only way two machines with the same version of 4GL would use different compilers would be if somebody doctored the "c4gl" script. >On machine A, programs work when compiled with cc or rcc. >On machine B, programs work when compiled with rcc, yet compilation with >cc rounds float variables to whole values, i.e 10.65 = 10.00 Sounds very odd. Perhaps machine B's cc uses shared libraries, but the libraries are actually rcc libraries. Back in the 4.10 days, we were strict about using rcc (the portable, bundled compiler, essentially the same as cc on Interactive or AT&T 3.2), since it minimized the differences between SCO ports and other UNIX SVR3.2 ports. Later, I heard that someone decided to use "cc" (the Microsoft-based compiler) instead... I was out of that division by then, so I never heard the reasoning why. >If this is the case, why does cc cause rounded errors? >enlighten me, please? Again, best guess I have is that the run-time libraries are incompatible with the compiling system. You might try coercing the linker to static linking and see what happens (performance could be adversely affected). >BTW Flags on rcc are -Dunix -Di386 Those are built-ins anyway, in my experience. -- (c) 1996 Alan Denney aland@informix.com On the 12/1/95 "NewsHour with Jim Lehrer", a story about Whitewater included an interesting close-caption error: the name "Madison Guaranty" was close-captioned as "Greg Maddux".