Re: [open-cobol-list] X86_64 branch

[David Black]

Patrick,

The work performed developing OpenCOBOL is a masterpiece of development. 
It has taken many man years of very dedicated people to create it. 
Keisuke, Roger, Brian and many others have spent years of their 
available time making it happen.

They are the first ones to do it successfully without the funding and 
support of a hardware or software company behind them. They had the 
vision many years ago and now it is real and usable. I know that they 
have more things that they want to add on and they, and the community 
will make it happen.

Based on my experience with COBOL over the years, and knowing most of 
the original COBOL compilers author personally, the one thing they had 
in common was that they lived, breathed, and thought in the computer 
language they were writing the compiler in for many years. The compilers 
written in assembler, or some language unique to the hardware using 
hardware features. They had to support the COBOL standards and 
portability of the compiler was not even a consideration. The compiler 
on the IBM 7070 was completely different than the one for the IBM360 and 
also both were different that the one for the IBM series 1. IBM had a 
different compiler for each hardware platform. That is a lot of 
different compilers to write and support.

Unix and Linux are standard in almost all the features that are used in 
COBOL so a program in C would run in almost all cases with out a 
problem.  Cygwin is used to solve the problem of running the C code on 
Windows. Other cases where authors of COBOL that wanted to support 
Oracle, a pre-processor (pro*COBOL) is used which converts the SQL 
statements to blocks of COBOL code. That output from the pre-processor 
is compiled with the normal COBOL compiler.

There are only two philosophies to building a compiler. The first it is 
tailored to the uniqueness of the hardware it is being written for. Or 
second is written in some form of "P" code that is 
interpreted/translated at run time with a run time program. That program 
is tailored to a particular brand of hardware and supports I/O and other 
unique features on that hardware. In Java you have a runtime program, 
for each brand of hardware, that executes the Java code. That is the 
method that AcuCOBOL used and a number of other software houses used. 
IBM, UNISYS, HP and others wrote their compilers using an army of people 
and implementing feature above the standards of COBOL.

My 5 cents worth is, that if want your COBOL code to run on another 
brand of hardware or operating system, that you write a pre-compiler 
that will generate new COBOL code by write replacement code that can be 
compiled by OpenCOBOL into C that will run on your target machine. Then 
you do not need to become an expert in C and need to only know, 
in-depth, the feature that you want to implement on the target machine. 
Writing a pre-processor is far easier that forking OpenCobol, mastering 
C and understating the source of OpenCOBOL.

Best of luck on what ever route you take for solving your portable problem.

David Black


On 2/24/2014 1:28 PM, address@hidden wrote:
> Re: X86_64 branch