Subject: Re: CVS commit: src/sys/arch/alpha
To: Jason Thorpe <>
From: Stephen Degler <>
List: source-changes
Date: 09/19/2003 18:59:05
Hash: SHA1


Of course gdb53 will work fine with software single stepping.  The config d=
get it wrong without intervention. ./configure + make doesn't compile with=
gcc3.3.1 + current.

I think the most compelling argument you made was that gdb should know when=
the breakpoints are set.  I agree with this, and hadn't thought about that=
when I was considering how best to move forward.

Since some of the problems vs gcc3.3.1 are in /usr/src/gnu/dist/gdb, I gues=
I'll be sending them to you to  be fed back to  the FSF?

I'll back the PT_STEP change out some time over the weekend.


On Friday 19 September 2003 02:03 am, Jason Thorpe wrote:
> On Thursday, September 18, 2003, at 12:42  PM, Stephen Degler wrote:
> > gdb53 expects *BSD to have the support, and miscompiles if it doesn't.
> >  I do
> > agree with you completely that it is not strictly necessary.
> I don't believe that gdb53 expects NetBSD/alpha to have PT_STEP
> support.  Note also that I am the maintainer of NetBSD host/target for
> FSF GDB, and I'm pretty sure I made the software single-step work in
> GDB 5.3 for NetBSD/alpha.
> > However, it does make NetBSD more consistent with the other BSD's and
> > less of
> > a special case.  Its was for that reason that I thought that it would
> > be best
> I don't particularly care if it makes NetBSD more consistent with other
> BSDs in this regard.  It's simply incorrect to have PT_STEP for the
> Alpha, and I would like this change to be backed out.
> > to include the support.  I'd think that we'll end up with less
> > maintenance if
> > we are less different.
> >
> > And of course, I will be clearing any future alpha commits with you.
> Thank you.
>          -- Jason R. Thorpe <>
Version: GnuPG v1.2.2 (NetBSD)