Publishing binary interfaces [was Re: [PATCH] Move "types deeply equal" code from py-type.c to gdbtypes.c]

[Doug Evans <dje@sebabeach.org>]

Stan Shebs <stanshebs@earthlink.net> writes:

> The comments should be maybe be phrased to emphasize the
> quasi-librariness of the code, so that future generations aren't
> confused if it seems unused by core GDB (which could happen if python
> becomes plugin
> instead of subdir, for example).

btw ...

While I'm all for publishing gdb as, in part, a collection of libraries,
and thus publishing binary interfaces, one reason why Python was added
the way it was is so that we can make gdb extensible without having to
publish a binary interface (we export some enums to Python, but that's
small potatoes compared to a real binary interface).
Some people IIRC were *profoundly* against publishing binary interfaces.

Is the community changing it's mind on binary interfaces?

Do people actually envision dlopen'ing GDB's Python extension?