Any objections to making the gnulib <sys/wait.h> apply the gcc warn_unused_result attribute to waitpid(), even though glibc does not (yet) do likewise? After all, if you fail to check for failures on this syscall (particularly EINTR failures, where you should normally repeat the waitpid() call), you can end up leaking unreaped child processes, which is a rather nasty resource leak to debug.

I ask this because Coverity was complaining about a couple places where libvirt was calling waitpid() and ignoring the return value.

--
Eric Blake   ebl...@redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

Reply via email to