Re: [RFA]: Watchpoints per thread patch

[Daniel Jacobowitz <drow at false dot org>]

On Thu, Oct 28, 2004 at 03:47:22PM -0400, Jeff Johnston wrote:
> Daniel Jacobowitz wrote:
> >On Wed, Oct 27, 2004 at 07:17:13PM -0400, Jeff Johnston wrote:
> >
> >>Were you thinking of add_thread()?  If so, we would have to move the 
> >>calls to add_thread so they never occur before an attach because the 
> >>low-level observers will need the thread already attached.
> >
> >
> >Oh, that's a good point.  Do you think that's a reasonable change to
> >make?
> >
> 
> It is a can of worms.  I can move the add_thread call in attach_thread 
> easily enough, but there are other calls to add_thread strewn about.  For 
> example, corelow.c calls add_thread as does infrun.c when it finds a new 
> process.  I certainly don't see it being valid for either of these 
> scenarios to insert/remove all watchpoints.  My personal preference would 
> be to leave it where it is for now.

There are two separate questions here:
  - When do we need to be adding and removing watchpoints from threads
    on GNU/Linux?
  - When should an observer named "new_thread" be called?

If it's not valid to do the former action at all the latter points,
then it's not the right observer to be using.

The code in infrun is never reached for native GNU/Linux threads, btw;
I'm not sure which targets if any do reach it.  I don't believe remote
GNU/Linux threads do either.

-- 
Daniel Jacobowitz