This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [0/6] breakpoints_ops for all kinds of breakpoints
Pedro Alves <pedro@codesourcery.com> writes:
> On Friday 22 July 2011 18:32:59, Phil Muldoon wrote:
>> I think it is great. It clears up several (dozen) ambiguities with
>> breakpoint operations. I don't think this will affect the Python
>> breakpoint_ops work (other than some refactoring, but we'll come to that
>> bridge when your work is checked in).
>
> Yeah, I don't think your current patch would need more than a
> some trivial updates, but then again, I did leave the print_one
> method behind in this conversion, which was the method that
> I raised having the issue that doesn't seem to fit for gdb's own
> breakpoints currently, so we should clean that up before
> considering exposing it as public API.
I have been thinking on your comments quite a bit in the last few
weeks, and about external facing APIs in general. I don't think it is
sensible to make 1:1 mappings of these breakpoint_ops APIs
to Python. I think "print_one" "print_it" mean nothing to a Python
scripting user. I've come to the conclusion that we should abstract
these functions to concepts a Python scripter can better understand. So
"print_one" will not be directly exposed as before. The Python API,
will end up calling print_one, sure, but there will (I hope) be
sufficient levels of abstraction that any future refactoring will be
coped with internally. We already massage the data to-and-from GDB
extensively anyway, so it makes better sense to the user.
So my plan is to submit a patch after yours is checked in, with the new
APIs. We can look again, then, if it makes sense at that point, and if
any future refactorings of print_one will really matter to the user from
an API perspective.
Cheers,
Phil