dlmopen and core dumps

[Carlos O'Donell]

On Fri, Dec 14, 2012 at 12:25 PM, Pedro Alves <palves@redhat.com> wrote:
> Here you're ignoring my point (all I had said expect the parenthesis), about
> other tools, etc.

I'm not ignoring it. I honestly think that no matter what happens
we'll eventually
end up wanting some hybrid of declarative and procedural. You disagree with
that.

What's wrong with the other tools linking in a python interpreter and running
the appropriate procedure to get information out of the glibc data they have
in their buffers?

The problem I have with declarative languages is that we will need to start
making symbols for *everything* e.g. lma, vma, base of X, address of Y,
and using labels for those things to talk about other things.

For each of these labels gdb will have to know how to provide the right
value for the declarative information to be interpreted correctly. That's a
pretty big burden.

I've seen some debuggers with declarative definitions of things e.g.
RTOS data structures, and Linux kernel data structures.

There is probably nothing you can't do with a declarative definition, since
in the abstract it's a generic program that gdb runs to get an answer.

Why not just run a generic program, or rather a python script?

Cheers,
Carlos.