This is the mail archive of the
gdb@sources.redhat.com
mailing list for the GDB project.
Re: remote.c's "async", "extended-async" back ends
- To: jtc at redback dot com
- Subject: Re: remote.c's "async", "extended-async" back ends
- From: Andrew Cagney <ac131313 at cygnus dot com>
- Date: Mon, 05 Feb 2001 10:04:38 -0800
- Cc: gdb at sourceware dot cygnus dot com
- References: <5mbsslsw3b.fsf@jtc.redback.com>
"J.T. Conklin" wrote:
>
> I noticed that if we adopted the step over range command, I'd have to
> add what I've done to remote_resume() to remote_async_resume(). This
> is not a problem of course.
>
> I seem to recall that the goal is to merge (or perhaps replace) the
> async remote back ends with the sync versions. What must be done in
> order to move this forward? As one of the remote.c maintainers, I'd
> like to get rid of this largely duplicated code.
>
> I recently objected to removing the sync event loop. If this is the
> holdup, I'm willing to spend some time to solve the problem we have
> with stopping the target when the async event loop is active.
See the TODO file. There is a really nasty road block in how GDB
connects to the target. Briefly the ``target xyz abc'' goes straight
down through to the target vector - GDB's kernel gets no say in what
happens.
The ``target xyz abc'' command should go to GDB's kernel and then GDB's
kernel should invoke the correct target methods.
Andrew