Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!husc6!seismo!ll-xn!cit-vax!news
From: news@cit-vax.UUCP
Newsgroups: comp.lang.c++
Subject: Variable constants
Message-ID: <2032@cit-vax.Caltech.Edu>
Date: Fri, 13-Mar-87 09:31:21 EST
Article-I.D.: cit-vax.2032
Posted: Fri Mar 13 09:31:21 1987
Date-Received: Sat, 14-Mar-87 10:49:31 EST
Reply-To: jon@oddhack.UUCP (Jon Leech)
Organization: California Institute of Technology
Lines: 43

Organization : California Institute of Technology
Keywords: 
From: jon@oddhack.Caltech.Edu (Jon Leech)
Path: oddhack!jon

	While writing some C++ code which was intended to call FORTRAN
routines, I was using constant-dimensioned matrices. The logical thing
to do, then, was

	const int n = 20;
	double m[n][n];

	The FORTRAN routines wanted the matrix and the dimension of
the matrix passed, along with 6 other parameters. Initially I didn't
feel like templating the routine (pretty difficult to do correctly anyway, 
given that FORTRAN does not distinguish between array dimensions), so I 
just had:

	extern void foo(...);
	foo(m, &n, etc...);

	All fine and dandy. Now I decide to add a new array, declared
AFTER the subroutine call:

	double x[n];

	Oops! The enlightening error message 'error: cannot
evaluate constant' occurs. This was most confusing. After I bit I
decided this meant cfront thought that 'n' might be modified
within 'foo'. OK, so I proceeded to template the function:

	extern void foo(double m[n][n], const int *pc, etc...);

	Unfortunately, the same error message still occurred (this is
1.1, incidentally). I consider this a bug. The template is clearly
specifying that whatever is passed is not being modified, yet cfront
appears to believe this anyway. Perhaps Bjarne would care to
comment? ``Fixed in 1.2'', or must I work around it, or am I just
missing something obvious (it is 6:30 AM, after all)?

    -- Jon Leech (jon@csvax.caltech.edu || ...seismo!cit-vax!jon)
    Caltech Computer Science Graphics Group
    __@/