Ordering problem when using Ruby popen2 to interact with C++ iostreams - ruby

Question

I'm using Open3's popen2 to interact with a simple C++ program's iostreams. My understanding is that std::cin and std::cout are independent, but the order in which I have my popen2 block's IO objects read/write calls seems to make a difference. My C++ program is:
int main(int argc, char** argv) {
std::string input;
std::cout<<"EXECUTE TASK"<<std::endl;
std::cin>>input;
std::cout<<"END"<<std::endl;
}
My ruby script is:
require 'open3'
expected_string = "EXECUTE TASK"
Open3.popen2('~/Sandbox/a.out') { |stdin, stdout|
stdin.write('\n')
stdin.close
results = stdout.readlines
puts results
}
The above works fine, but if I move the stdout.readlines before the stdin.close, the ruby script will hang. My intent is to conditionally write the \n to stdin if the C++ program writes expected_string to standard out first, but I'm forced to close the stdin stream before I can execute readlines. Like I said, my understanding was that the two streams are independent, and the file descriptors returned by popen2 appear to be independent as well, so why would the order matter?
Any help is appreciated. Thanks.
Solution with full scope of what I was trying to accomplish (someone may find this helpful):
int main(int argc, char** argv) {
std::string input;
std::cout<<"1"<<std::endl;
std::cout<<"2"<<std::endl;
std::cout<<"3"<<std::endl;
std::cout<<"4"<<std::endl;
std::cout<<"5"<<std::endl;
std::cout<<"EXECUTE TASK"<<std::endl;
std::cout.flush();
std::cin>>input;
std::cout<<"END"<<std::endl;
}
require 'open3'
expected_string = "EXECUTE TASK"
Open3.popen2('~/Sandbox/a.out') { |stdin, stdout|
found = false
begin
while(result = stdout.readline)
puts result
if(result.include?(expected_string))
found = true
break
end
end
rescue
raise "Exception caught while reading lines"
end
stdin.write('\n')
stdin.close
}

Answer 1

Looks like a deadlock:
The ruby script is calling IO::readlines, which will not return until the entire stream has been read.
The C++ program does not actually terminate until it receives a carriage return.
You may want to call IO::readline, which will return each line as it is received, or reorder the two scripts so that there is not a deadlock.

Answer 2

It is highly likely that you just need stdout.flush after anything that you expect your correspondent to act on.
Because the pipe set up by #popen is not a tty, it will not default to line buffering. It will do block buffering. You will need to force your stream to act in datagram fashion by calling #flush on "record" boundaries.
Also, see IO#sync= for a way to flush all I/O automatically.