On Wed, 02 May 2018 01:03:39 +0200, Jan Kratochvil wrote:
> On Thu, 26 Apr 2018 22:12:16 +0200, Jan Kratochvil wrote:
> > On Fri, 20 Apr 2018 16:49:39 +0200, Pedro Alves wrote:
> > > ~~~~~~~~~~~~
> > > Previously, when the hardware reported a watchpoint hit on an address
> > > that did not match our watchpoint (this happens in case of instructions
> > > which access large chunks of memory such as "stp") the process would
> > > enter a loop where we would be continually resuming it (because we did
> > > not recognise that watchpoint hit) and it would keep hitting the
> > > watchpoint again and again. The tracing process would never get
> > > notified of the watchpoint hit.
> > > ~~~~~~~~~~~~
> > > 
> > > ... I'm left with the impression that ADDR_TRAP could be even
> > > lower than addr_watch_aligned, in which case we'll still miss
> > > watchpoints.  I wondering whether GDB should be using a similar
> > > trick as that kernel patch does.
> > 
> > This is new for me what you found.  I just did not expect the changed region
> > region could be larger than aligned 8 bytes.
> > 
> > Unfortunately I cannot reproduce that so I cannot do much with that.
> > Does anyone know how to reproduce it?
> 
> Attaching the change I made, I will repost the patch in a next mail.

Here is the new patch, if that new testcase
	PASS: gdb.base/watchpoint-unaligned.exp: size8twice write

fails for anyone I can check how to avoid it.


Thanks,
Jan