[PATCH] test-skeleton: Kill any child process's offspring

[Roland McGrath]

> > setpgid should not fail.  Checking that it actually doesn't is just good
> > practice.
> 
>  But that's a preexisting problem, not relevant to my proposal.  And 
> therefore if you think it needs to be addressed, then I think it should be 
> corrected as a separate change.  Please feel free to make it at any time.

No.  Your change presupposes that setpgid did not fail, which was not
presumed before.