From: lynnmcguire5@gmail.com   
      
   On 10/26/2024 6:51 AM, Thomas Koenig wrote:   
   > Lynn McGuire  schrieb:   
   >> On 10/24/2024 1:28 AM, Thomas Koenig wrote:   
   >>> Lynn McGuire  schrieb:   
   >   
   >>>> F2C fixes the other big problem   
   >>>> automatically, the change of initial array index from one to zero.   
   >>>   
   >>> If I remember correctly, it does so by issueing invalid C (or   
   >>> C++), by using negative offsets from pointers.  Might work now,   
   >>> might not work tomorrow.   
   >>>   
   >>> But note the IIRC above.   
   >>   
   >> I want to move to a monolanguage environment.  50,000 lines of my   
   >> calculation engine are C++ already.  850,000 lines to go.   
   >   
   > That motivation, I understand, especially if the GUI code is in C++,   
   > but there is a caveat.  Consider   
   >   
   >        subroutine foo(i,n)   
   >        integer array(10)   
   >        common array   
   >        integer n   
   >        integer i(n)   
   >        integer k   
   >        do k=1,n   
   >           i(k) = k + array(k)   
   >        end do   
   >        end   
   >   
   > which gets translated by stock f2c (to which you may have made   
   > adjustments) into   
   >   
   > #include "f2c.h"   
   >   
   > /* Common Block Declarations */   
   >   
   > struct {   
   >      integer array[10];   
   > } _BLNK__;   
   >   
   > #define _BLNK__1 _BLNK__   
   >   
   > /* Subroutine */ int foo_(integer *i__, integer *n)   
   > {   
   >      /* System generated locals */   
   >      integer i__1;   
   >   
   >      /* Local variables */   
   >      static integer k;   
   >   
   >      /* Parameter adjustments */   
   >      --i__;   
   >   
   >      /* Function Body */   
   >      i__1 = *n;   
   >      for (k = 1; k <= i__1; ++k) {   
   > 	i__[k] = k + _BLNK__1.array[k - 1];   
   >      }   
   >      return 0;   
   > } /* foo_ */   
   >   
   > The common block handling looks OK, but the dummy argument   
   > (aka parameters, in C parlance) handling is very probably not.   
   >   
   > The "parameter adjustment" above is explicitly listed as undefined   
   > behavior, in annex J2 of n2596.pdf (for example):   
   >   
   > "Addition or subtraction of a pointer into, or just beyond, an   
   > array object and an integer type produces a result that does not   
   > point into, or just beyond, the same array object (6.5.6)."   
   >   
   > Undefined behavior is the worst kind of error in your program   
   > that you can have in C, it is not required to be diagnosed, and   
   > compilers can, and do, make optimizations based on the assumption   
   > that it does not happen, so this is liable to break in unforseen   
   > circumstances.   
   >   
   > So if your version of f2c does the same, I would check the C++   
   > standard if if has a similar provision (I strongly suspect so,   
   > but I don't know), and, if that is the case, modify your version   
   > of f2c to generate conforming code for array dummy arguments.   
   > Otherwise, you are betting your company.   
      
   First, I include all of my 300+ common blocks as 200 files.  I converted   
   those separately and cleaned them up so that the static variables and   
   defines are easy to peruse and understand.  I delete all of the local   
   common block conversions by f2c in the subroutines and change them back   
   to include files.  An easy cleanup that I have to do 5,000 times (4,000   
   to go now plus the 100+ subroutines that we have modified for customers   
   since I started the conversion project two years ago).   
      
   I also removed the parameter adjustments from my copy of f2c.  It is a   
   little tricky but as you say, they are not legal code in C++.   
      
   I have extensively modified my copy of f2c so that it generates better   
   legal C++ code and many other improvements.  But my changes are subject   
   to the old 80 / 20 rule.  I have 80% automated conversions and 20% hand   
   conversions.  I do have a little experience here: Fortran since 1975, C   
   since 1987, and C++ since 2000.   
      
   Thanks for the warnings,   
   Lynn   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)