Executive summary: I need a way to determine whether a Windows process I've spawned via _spawnl and am communicating with using FDs from _pipe has died.
Details:
I'm using the low-level CRT function in Windows (_eof, _read) to communicate with a process that was spawned via a call to _spawnl (with the P_NOWAIT) flag. I'm using _pipe to create file descriptors to communicate with this spawned process and passing those descriptors (the FD #) to it on the command line.
It is worth mentioning that I don't control the spawned process. It's a black box to me.
It turns out that the process we are spawning occasionally crashes. I'm trying to make my code robust to this by detecting the crash. Unfortunately, I can't see a way to do this. It seems reasonable to me to expect that a call to _eof or _read on one of those descriptors would return an error status (-1) if the process had died.
Unfortunately, that isn't the case. It appears that the descriptors have a life of their own independent of the spawned process. So even though the process on the other end is dead, I get no error status on the file descriptor I'm using to communicate with it.
I've got the PID for the nested process (returned from the _spanwnl call) but I don't see anything I can do with that. My code works really well except for one thing. I can't detect whether the spawned process is simply busy computing me an answer or has died.
If I can use the information from _pipe and _spawnl to determine if the spawned process is dead, I'll be golden.
Suggestions very welcome.
Thanks in advance.
UPDATE: I found a fairly simple solution and added it as the selected answer.