[prog] compiling (or failing to) old fortran on linux

[Shirrell]

Conor Daly wrote:

>Hi all,
>
>I have a (biggish) number of fortran programs on a solaris machine that I
>want to compile on a linux system.  There is the added complication of some
>embedded SQL routines so a precompiler parses the fortran first.  The
>sequence of the solaris machine is:
>
>esqlc <source>.sf
> (produces <source>.f
>
>f77 -fast -xO5 <source>.f /users/ingres/lib/*.a -lelf -lm -lnsl -lsocket
>
>and all works fine.  However, when I try to compile on an RH9 box using:
>
>esqlc <source>.sf
> (produces <source>.f
>
>g77 $user.f -L${II_SYSTEM}/ingres/lib -lingres -lm -ldl -lrt
>
>I get the following errors:
>
>           type 935,razer
>           ^
>Unsupported VXT statement at (^)
>  
>
Every Fortran has its own little extensions and Solaris is on the 
permissive side
with some early f90 features.  g77 tends to stick to the standard.

>
>   2027    format (x,i2,i10,i10,' Smd:',i7,' ','Diff: ',i5,
>  
>
While I can't be much help the  '(x,' without a size specifier such as
1x is certainly not vanilla Fortran and I have
found that g77 is less permissive that f77.  

>                   ^
>Invalid X specifier in FORMAT statement at (^) -- correct form: nX
>
>           if((ans.eq.'n').or.(ans.eq.'N')) then
>                  1   2
>Equality operator at (1) must operate on two scalar (not array)
>subexpressions, two function invocations returning arithmetic or character
>scalars, or a combination of both -- but the subexpression at (2) is an
>array
>
How is ans declared -- this looks ok if it is something like   
character*n ans

>
>             if (zwdif.lt.0) signw = '-'
>                             1       2
>Type disagreement between expressions at (1) and (2)
>
>  
>
Again, how is signw declared: If it is character*n it should be ok on 
g77.  I don't
think switches will help.  We move Fortran code from solaris to g77 all 
the time
and compile without switches.

>Does anyone know the correct switches for g77 to make this stuff compile?
>(-fvxt -fvxt-intrinsics-enable) doesn't seem to make a difference...
>
>Thanks,
>
>Conor
>
Regards,
Shirrell at pstat.com

>  
>