eliminate deprecated_insert_raw_breakpoint. what's left.

Peter Schauer peterschauer@gmx.net
Tue Sep 9 00:25:00 GMT 2014


> 
> On 09/08/2014 10:34 PM, Joel Brobecker wrote:
> 
> > I was looking at how to replace that call, but I am not sure
> > how to fix the code up, though. Perhaps we could just write
> > the breakpoint instruction in by hand, rather than go through
> > the breakpoint module? After all, it is already doing almost
> > everything else by hand!
> 
> Indeed.
> 
> > In fact, looking at the code again now, I'm a little more tempted
> > to see what happens if we remove it ;-).
> 
> Me too.  And seriously.  :-)
> 
> I traced it back to I think the original rs6000 port, in 1991...
> 
> commit 41abdfbd2de07837ba8088092765154eaa66351d
> Author: John Gilmore <gnu@cygnus>
> Date:   Tue Nov 12 15:50:47 1991 +0000
> 
>     * rs6000-pinsn.c, rs6000-tdep.c, rs6000-xdep.c, tm-rs6000.h,
>     xm-rs6000.h:  New files.
>     * xcoffexec.c:  New file for handling AIX shared libraries.
> 
> 
> 
> We already see this then, in rs6000-xdep.c:
> 
> +       /* execute one dummy instruction (which is a breakpoint) in inferior
> +          process. So give kernel a chance to do internal house keeping.
> +         Otherwise the following ptrace(2) calls will mess up user stack
> +         since kernel will get confused about the bottom of the stack (%sp) */
> +
> +       exec_one_dummy_insn ();
> 
> 
> This sounds like working around a (very) old kernel bug...

No, not a bug, but consequences of a very peculiar ptrace implementation,
at least up until and including AIX 4.3.

See my other post for a more detailed explanation.

> I can't believe anything resembling a modern system would need
> such a monstrosity!  :-)  I vote just removing all that.

Me too, provided that newer AIX versions have a better ptrace implementation,
making this  monstrosity unnecessary.

-- 
Peter Schauer			Peter.Schauer@mytum.de



More information about the Gdb-patches mailing list