Re: Removal of VAX/VMS support

[Ian Lance Taylor <ian at airs dot com>]

Jim Blandy <jimb@redhat.com> writes:

> For GDB, marking ports as obsolete and eventually deleting them has
> the advantage that clients of old interfaces to core functionality
> gradually disappear.  Once they're all gone, you can delete whatever
> cruft in the core was supporting the old interfaces, which often frees
> you up to do better things with the core.  I can't wait until the
> clients of the old stack unwinding interfaces are gone from GDB ---
> that stuff is hard to reason about, and easy to use wrong.  But
> because Andrew goes around threatening to delete ports if they don't
> get reworked to use the new interfaces (that's not quite an accurate
> description of the tactic, but it's something like that), they will
> eventually no longer be present to confuse people.

I think it's worth pointing out that gdb is quite a bit more
complicated than the binutils.  The binutils are all conceptually
quite simple.  gdb is tougher because it has to look at the code in an
unnatural way--pulling it backward from executable to source.  The
binutils all go in the natural direction, from source to executable
(except for objdump --disassemble or --debugging, I suppose).

The conceptual complexities people encounter with the binutils are
mainly a matter of understanding the somewhat inappropriate design and
the very confusing implementation, not the actual task which the
program is supposed to do.

Ian