On Friday 15 January 2010, Igor Skochinsky wrote:
> This is pretty standard. Most targets cannot continue from a
> breakpoint (i.e. they just stop at the same position), so GDB removes
> breakpoint, steps, puts it back and then continues.
> On the other hand, if there's no breakpoint at the current address, GDB
> should just issue continue straight away. If it doesn't, there might
> be something else at work...

Another theory about what might be happening on ARM11:

There's partial hardware breakpoint support in the new DMP code,
and maybe it's kicking in when it shouldn't.  It *should* kick in
to erase breakpoints at OpenOCD startup; they'd be left over from
a previous debug session.

But unless the particular core is using the DPM breakpoint
code, it shouldn't kick in at any other time.

I might not have noticed such a bug when I tested a while back,
since I have (still incomplete) patches to make ARM11 use that
breakpoint code.  But mainline should still be using the older
ARM11-internal breakpoint code (which among other things does
not support soft breakpoints).

I'll look at that as soon as I finish a few other bits; I've
not yet done my ARM11 sanity testing for the 0.4 code.

- Dave

_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to