Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!ucsd!pacbell.com!ames!haven!uvaarpa!murdoch!news From: randall@Virginia.EDU (Ran Atkinson) Newsgroups: comp.os.msdos.programmer Subject: Re: TurboC and TurboC++ Message-ID: <1991Jan8.195900.28885@murdoch.acc.Virginia.EDU> Date: 8 Jan 91 19:59:00 GMT References: <350bahrd@yoda.byu.edu> Sender: news@murdoch.acc.Virginia.EDU Reply-To: Ran Atkinson Followup-To: comp.os.msdos.programmer Distribution: comp Organization: University of Virginia Lines: 51 In article <350bahrd@yoda.byu.edu> bahrd@yoda.byu.edu writes: % I was wondering what the difference between TurboC 2.0 and TurboC++ % 1.0 in the standard C mode was? % Are there any portability porblems if I write code in TurboC++ % using standard C conventions and then giving the code to a friend % who has Turbo C 2.0? Please let me know!!! TC 2.0 was close to complying with ANSI X3.159 C but not quite there. TC++ 1.0 is a lot closer to being ANSI X3.159 compliant. Most C programs will compile fine with TC++ 1.0 (all of the ones I had lying around just compiled with an occasional new warning but no errors). If you write code that depends on some really fine point of ANSI C then you might have problems with it on TC 2.0 but most programs I've seen don't have this problem. In article Dion Hollenbeck writes: > Unfortunately I do not know specifics, but I just talked to Borland > Customer Service about upgrading from TC to TC++ and one question > that I asked was "Will my C programs just compile without change?" > and they informed me that I would have to PORT them. ALL of the C code I had working cleanly under TC 2.0 compiled as is without any code changes for TC++ 1.00. There was no "porting" effort at all. Apparently Borland is trying to be safe in what they claim. Borland has repeatedly stated that they don't plan to have any new version of TC in the near future and specifically told me that TC++ was there development focus and is the direction to move in. On another note, tests I did this fall with TC++ and Zortech C++ for MSDOS using C++ programs (that use inheritance) showed that with minor source code changes in member functions, either Zortech or Borland could generate the faster code or the smaller executable (sans debug records). This was done in part with proprietary code so I can't post the code. It is consistent with earlier results published in _Computer_Language_ earlier in 1990 with regard to C (not C++) compilers. That article included code fragment examples for C which give some indication of what I mean by "minor" changes. There are a lot of factors to consider in selecting a compiler, but increasingly I am convinced that optimisation is not greatly different among the main compiler vendors (Borland, Zortech, Watcom, MS) and that support, library quality, and such are much more important and are significantly different with different vendors. I have no connection with any compiler vendor other than as an ordinary customer. Opinions expressed are the author's and aren't necessarily those of the University of Virginia.