Hi Jim, > I think it is uncontroversial. > > Pushed. I hope that's ok.
The time window you gave me to react on this one was less than 24 hours; I couldn't answer in time. On this one, I would have preferred to silence the warnings in a cheaper way. It's irrelevant for the train of events inside the test whether its diagnostic output on stdout has really reached its destination. Too many ASSERTs, especially on irrelevant details, can be confusing. Of course I agree to insert ASSERTs on events that do influence the test, like ASSERT (pipe (fd) == 0); in your other patch. If pipe(fd) failed, it's better to see a message sooner than later. Anyway, not a big deal. Bruno
