 |
 |
> Date: Sat, 29 Nov 1997 18:47:44 -0500 > From: "Simon G. Smith" <[log in to unmask]> > Hey, I have all the love in the world for Fortran, but since when is C > "not highly portable"? I'm monitoring a project to add F90 features to > a F77 code, and I'm seeing all kinds of compiler dependent results, and > not in the libraries (C's weakness) but rather in the language features > (derived types, pointers, allocatable arrays, and the combination of > these). I would argue that these usually aren't language problems. Most of the time thaty are either compiler bugs or non-standard coding practices. If you'd like to list features of the language that are compiler-dependent, I, for one, would stop using those features. I've run into a fair number of portability problems, and only two I can recall immediately are truly language problems, and one is arguably a "quality of implementation" problem and the vendor promised it was fixed in the next release. > I'm forced by management to use F90, so our project will make the best > of it, and take advantages of F90's strengths. However, I think I'd > have a much more portable code using ANSI C. One thing to keep in mind is that Fortran90 compilers are, on aaverage, newer than ANSI C compilers, and aren't all 'up to snuff' yet. I've run into very few C programs (ANSI or otherwise) that compiled and ran identically on different platforms without any porting effort. Recently, the GNU `configure' program has helped a great deal on this front, but the effort to make a C program `configure'able can be substantial. I have no idea if `configure' can be used with Fortran programs, although I do have experience trying to use `cpp' on different platforms with Fortran code and it is a pain, and `cpp' is at the heart of `configure'. > Simon David %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%