Vladimir Prus wrote:

> Daniel Jacobowitz wrote:
> 
>> On Wed, Apr 23, 2008 at 08:45:54PM +0100, Pedro Alves wrote:
>>> I've tested this patch on top of Vladimir's and with displaced
>>> stepping on, on x86-pc-linux-gnu.  Luis tested it on PPC, where
>>> it also fixed things.
>> 
>> OK when the original patch is committed.
> 
> Well, I have an obvious adjustment to the original patch to fix this
> issue in a more clear way -- my making should_insert_location not
> return false for an already inserted location.

I've checked in the below. I've tested this with default mode and
in always-inserted mode, but not in async mode or in async mode +
displaced patch. I'll do that now and should there be issues,
post them separately -- it seems more productive than having folks
send patches against not-yet-checked in code :-)

- Volodya