Roland Spatzenegger <cymric@npg.net> writes:

> I have a performance problem in my linux-application which starts
> other programs with fork/exec.

You probably should be asking in c.o.l.d.{apps,system} ...

> In the C++ Application I use ACE/TAO (so the fork exec is called from
> a Corba Thread) and the stdin/stderr from the shellscript are
> redirected to a file (in the fork()).

Fork()ing mutithreaded apps is asking for trouble: it requires that
*all* libraries you link to implement atfork handlers correctly
(which may not be trivial).

> So I think the problem must be in the clone() or waitpid() call (?).
> Has anyone an idea how to track down the problem?

Try creating a minimal test case.
Also note that versions of kernel and glibc are likely to be
relevant (as well as whether you are using LinuxThreads or NPTL),
so you should state them.

Cheers,
--
In order to understand recursion you must first understand recursion.
Remove /-nsp/ for email.

[ Post a follow-up to this message ]