[NCLUG] COBOL Compiler for Linux?

[Anthony Foiani]

Stephen Warren <swarren at wwwdotorg.org> writes:
> That wouldn't make sense to me; C is basically as close to the HW as
> assembly for the right kind of C, which presumably any COBOL->C
> compiler would be producing. (So yes, going via C *could* reduce
> performance by generating bad C, but then again, there's no
> guarantee that a straight-to-binary compiler generates good code).

Hm.  That discussion has happened in the C++ community over the last
20-25 years, and there's good arguments and examples to be had on both
sides.

One notable feature is that the original C++ compiler ("Cfront") was
in fact a "C++ to C" compiler; according to Wikipedia, it kept up with
the language until they tried to add exceptions cleanly.  (Doing EH
with only a space cost and no runtime cost pretty much implies the
ability to muck around with the stack at will, which is difficult to
represent in portable C).

At least one company is still using the C++-to-C route, however:
Comeau C++.  Interestingly, it's broadly considered one of the most
standard-compliant compilers available.

Maybe COBOL is more strictly a subset of C than C++ is (isn't!), or
maybe the three decades of Moore's Law would allow for even awkward /
large / slow C versions of COBOL constructs to run "well enough".

t.

p.s. I remember when I got my 1GHz 1GiB PowerBook G4, and realized
     that I was carrying more raw compute power under my arm than my
     university mainframe had when I started college...