At 12:37 PM 1/30/2003, Bill Stoddard wrote: >This stuff kinda worked on Windows in the past. I need to dig some but I seem >to recall that it was basically impossible to do the exact same thing in >Windows as you do in Unix. The other_child_read in Unix will not (and cannot) >work the same way under Unix.
It's fine if they diverge when necessary. It's not fair to overload functions with two different meanings for two platforms :-) Let me play with the RegisterWait*() API in Win32 and see if we can't register that other child handle to invoke the registered callback upon death. We might even end up -cleaner- than the Unix implementation, without the need for extra polling in the MPM :-) Bill
