libthread_db thread handles
Daniel Jacobowitz
drow@mvista.com
Tue Jan 14 23:20:00 GMT 2003
On Tue, Jan 14, 2003 at 11:08:20PM +0000, Ulrich Drepper wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Daniel Jacobowitz wrote:
>
> > First of all, libthread_db doesn't necessarily create the
> > td_thrhandle_t; they're allocated by the caller of libthread_db, for
> > map_id2thr and map_lwp2thr. The iterators create them and then throw
> > them away, of course.
>
> ...which is why you cannot reply on the addresses of these objects to
> have any relevance. This is why I mentioned it.
OK. I understand; I don't expect the addresses to be relevant.
> > td_thrhandle_t is opaque; it is documented as opaque in <thread_db.h>.
> > I don't want to look at th_unique. Can I rely on the fact that the contents
> > of a td_thrhandle_t are stable and can be re-used if I save the
> > td_thrhandle_t?
>
> In the current implementation this is the case but I don't want to
> commit to this unless you can show me there are big benefits. It would
> help if you could describe what functions you see being used often and
> in which contexts. There are things in the libthread_db implementation
> which can be changed without breaking the ABI. We just need to know how
> you use the library.
To find the state of a thread, we need to first get a thread handle for
it and only then can we call td_thr_get_info. I'd like to save a copy
of the td_thrhandle_t when we get the TD_CREATE event, and be able to
hand it to td_thr_get_info until I see a TD_DEATH event (can a thread
die without us getting the TD_DEATH event? Is there some other thing
we should be watching for a thread to vanish?).
This can be a major timesaver, in addition to caching the LWP ID and
holding on to it as long as we know we've got a 1-to-1 mapping (via
not seeing any TD_SWITCHTO/TD_SWITCHFROM).
If you're not comfortable with it, let me know and I'll figure out
whether I can get the same effect another way.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer
More information about the Gdb
mailing list