This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: x86 incomplete/WIP software single-step implementation (Re: [PATCH] Do not respawn signals, take 2.)
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: palves at redhat dot com
- Cc: gdb-patches at sourceware dot org
- Date: Wed, 4 Jul 2012 21:44:55 +0200 (CEST)
- Subject: Re: x86 incomplete/WIP software single-step implementation (Re: [PATCH] Do not respawn signals, take 2.)
- References: <20120622145525.27114.25720.stgit@brno.lan> <20120622150456.GA5018@host2.jankratochvil.net> <4FE48B30.8070109@redhat.com> <4FEC9998.2040001@redhat.com> <4FF46296.9000706@redhat.com>
> Date: Wed, 04 Jul 2012 16:34:46 +0100
> From: Pedro Alves <palves@redhat.com>
>
> On 06/28/2012 06:51 PM, Pedro Alves wrote:
>
> > While I was waiting to get access to some Red Hat ARM machines,
> > I started cooking up a software single-step implementation
> > for x86, for easier-testing purposes. I though perhaps who knows
> > it might also prove useful in the future for something else, e.g.,
> > block-step fallback support.
> >
> > I got ARM access before I finished it, so it's largely incomplete (see
> > commit log), though it passes many tests. It's here in case it interests
> > someone:
> >
> > https://github.com/palves/gdb/tree/x86_software_single_step
>
> FYI, this is mostly complete and useable now. A few displaced-stepping
> regressions remain, and reverse debugging breaks (but ISTR that was a generic
> software single-step issue? If so, this should help debug it). I've also
> tested this with a hack that forces single-step for everything (including "continue")
> (there's a branch for that, see my github's wiki). This should make
> developing/testing all-stop-on-top-of-non-stop on software single-step
> archs a bit easier (for me).
I must admit I'd be reluctant to maintain i386/amd64 software single
step in GDB.