forums.silverfrost.com

[jjgermis]

I am using FTN95 to compile a relative old style fortran program. A few useful tips on this topic I found at:
1.) http://cs.ubishops.ca/ljensen/fortran/oldfort.htm and
2.) http://userpage.chemie.fu-berlin.de/diverse/doc/f77_info.html

I get several errors to that I am not familair with. Like the & in the parameter list from a subroutine, i.e.

[sparge]

Not sure about the first one because I don't understand what the & is doing in the source code myself, so if I weere a compiler I'd probably output much the same error message.

For the second one, I am (intelligently) guessing it implements some compiler-specific way of rendering a constant in hexadecimal notation, and FTN95 is objecting because its own specific syntax is slightly different. I think this will work:

CHARACTER*1 Z27,ZF9,ZF4
DATA Z27,ZF9,ZF4/Z'27',Z'F9',Z'F4'/

Andy

[PaulLaidler]

&9999 is probably intended to represent a label for an "alternate return".

This is not standard Fortran which uses a * not &.

[jjgermis]

I replaced & with * and it seems to solve the problem. Furthermore, I tried

[jjgermis]

If I have some code as given below, what does it mean. Hexadecimal seems to be some hint in the right direction. However, I have no clue what this should do.

[sparge]

[quote="jjgermis"]I replaced & with * and it seems to solve the problem. Furthermore, I tried

[jjgermis]

Thanks Andy! It compiles now. The program was developed when memory was a problem and a lot of data exchange is done via files (at least that what I assume). The given variables are only used as (probably) some specific format.

[jjgermis]

Maybe someone has experiecend the same: Three years ago (before my time here) it was recognised to update the compilers. For this several compilers were tested (my first choice would have been FTN95). From the tests FTN95 produced the most warnings and errors and based on this was not used.

I compiled the code from then to see what happens (using FTN95). I now realise that FTN95 discovered some errors which the other compilers allowed.

[JohnHorspool]

[jjgermis]

With the help from the Silverfrost forum I managed to compile the old style Fortran. In doing so I also discovered some errors in the code (good old FTN95). The first facelift was to change the (punched card style) input file (that has no text and only numbers) with a very simple XML-file. It might sound crazy, but associating software with XML already makes it modern! The next step would be to collect the user data using ClearWin+. I cannot wait to see the expressions of my colleagues when I inform them everything is done using FTN95

The XML-parser link: http://xml-fortran.sourceforge.net

[jjgermis]

I must admit that working through old style Fortran is sometimes not that easy. Especially the way do-loops, if-then and goto lables are implemented takes a while to get used to. However, there is also a solution available to this Beside the above mentioned link there is very useful software that I can recommend, i.e. http://www.scitools.com. I personaly like the graphical presentation of the calls from the main program.

It would be interessting to know what other possibilties are available to analyse source code.

[jjgermis]

Two useful converters from Fortran 77 to Fortran 90 are:
1.) to_f90.f90 by A. Miller and
2.) convert.f90 by Michael Metcalf
and can be found at http://www.unics.uni-hannover.de/zzzzwg1/frei.html

At the same link there is a reference to Salford/Silverfrost's Personal Edition

[PaulLaidler]

You could also try using the FTN95 command line option /CONVERT but this is rather limited.