Re: qL and qf remote packets [Re: [Kgdb-bugreport] Re: [discuss] kgdb-x86_64-1.6 for kernel 2.4.23]

[Daniel Jacobowitz <drow at mvista dot com>]

On Wed, Dec 31, 2003 at 02:25:42PM -0500, Andrew Cagney wrote:
> >On Tue, Dec 30, 2003 at 06:11:29PM +0530, Amit S. Kale wrote:
> >
> >>While q[s/f] packet itself doesn't have pid wrap-around problem, I can't 
> >>figure out what will happen to gdb's database of threads in following 
> >>scenario.
> >>
> >>1. GDB adds a thread with id 1500 to thread database.
> >>2. It finds that the thread has died later.
> >>3. Does it delete the thread from its database?
> >>4. It again finds a thread with id 1500 becase of wrapping around of pid. 
> >>If it has completely forgotten about previous thread in its dabase, there 
> >>shouldn't be any problem.
> >
> >
> >Indeed, there won't be any problem.  I believe that even if thread 1500
> >exists, and then dies and restarts between breakpoints, GDB still won't
> >get confused.
> 
> The user might (eventually).  They aren't going to be notified of thread 
> create/delete events.  Also, it could leave around per-thread 
> breakpoings no longer applicable to that thread.
> 
> However, until someone manages to present this as a real problem ...

I suggested a protocol for describing the create/delete events, but it
was dropped because it was "too asynchronous" IIRC?  Certainly with
kgdb stopping the system to report thread create/delete events isn't
acceptable, so I don't know quite what to do about that.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer