248 is a 'magic' ipx number?

Benjamin Saller Bender (case@loki.appliedtheory.com)
Tue, 20 May 1997 16:57:29 -0400

H. Peter Anvin <mailto:hpa@transmeta.com> writes:
>What appears to be the consensus is that with CLONE_PID, you can use
>either pid's or tid's where you currently use pid's. pid's and tid's
>will be allocated from the same number space. Using the pid will affect
>all threads, but using the tid will affect only one thread.
>
>There is no reason, then, ptrace() could not simply be called with the
>tid.
>

This is true. We could use ptrace in this way, but I think could and
should are two different things. There is another reason. ptrace is slow and
costly, requiring a number of context switches because the parent doesn't have
direct access to the child address space. If we every want to debug threaded
real time apps it seems like we will want as efficient a way of possible to
get at all the executable objects. I may be wrong, but Im not yet convinced
that a /proc based debugger wouldn't be better.

Thanks

-- 
Benjamin Saller Bender 			<case@AppliedTheory.com>
AppliedTheory Communications		Software Engineering Group
http://AppliedTheory.com/               Sentio aliquos togatos contra me conspirare.