Bye from forks sends signals?
Hey! Is this newsletter alive?:) Disclaimer: I'm not very familiar with system programming and POSIX I'm playing around with `native` and ZeroMQ and found a curious behaviour. (unless (fork) (wait 2000) (bye)) (setq Context (native "libzmq.so" "zmq_ctx_new" 'P)) (setq ZMQ_REP 4) (setq Socket (native "libzmq.so" "zmq_socket" 'P Context ZMQ_REP)) (native "libzmq.so" "zmq_bind" 'I Socket "tcp://*:") (buf Buffer 10 (prinl "Waiting for messages") (when (= -1 (native "libzmq.so" "zmq_recv" 'I Socket Buffer 10 0)) (prinl (pack "Error: " (errno) Basically, the main process sets up a server and waiting for a message and the child process simply waits for a bit and exits with `bye`. The `errno` is 4 (which is signal interrupt as i understand). The waiting in the child process is important because if it exits before zeromq code, everything's fine and the server is patiently waiting. My assumption here is that `bye` throws some signal? Why else would it affect zeromq in the parent process? Just looking for some explanation. Maybe even the proper way to resolve this. Actually, while writing this I found out about SIGCHLD which is apparently sent to parent on child's exit so I guess zmq_recv gets interrupted by that for some reason? Weird. Can anyone confirm that's what I'm seeing? P.S. completely offtopic but since I'm here. I just noticed that semicolons aren't treated as comments. Why? Can it be enabled? Otherwise my Emacs' lisp-mode comment/uncomment function is useless and no comment highlight either. -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
Bye from forks sends signals?
Hey! Is this newsletter alive?:) Disclaimer: I'm not very familiar with system programming and POSIX I'm playing around with `native` and ZeroMQ and found a curious behaviour. (unless (fork) (wait 2000) (bye)) (setq Context (native "libzmq.so" "zmq_ctx_new" 'P)) (setq ZMQ_REP 4) (setq Socket (native "libzmq.so" "zmq_socket" 'P Context ZMQ_REP)) (native "libzmq.so" "zmq_bind" 'I Socket "tcp://*:") (buf Buffer 10 (prinl "Waiting for messages") (when (= -1 (native "libzmq.so" "zmq_recv" 'I Socket Buffer 10 0)) (prinl (pack "Error: " (errno) Basically, the main process sets up a server and waiting for a message and the child process simply waits for a bit and exits with `bye`. The `errno` is 4 (which is signal interrupt as i understand). The waiting in the child process is important because if it exits before zeromq code, everything's fine and the server is patiently waiting. My assumption here is that `bye` throws some signal? Why else would it affect zeromq in the parent process? Just looking for some explanation. Maybe even the proper way to resolve this. Actually, while writing this I found out about SIGCHLD which is apparently sent to parent on child's exit so I guess zmq_recv gets interrupted by that for some reason? Weird. Can anyone confirm that's what I'm seeing? P.S. completely offtopic but since I'm here. I just noticed that semicolons aren't treated as comments. Why? Can it be enabled? Otherwise my Emacs' lisp-mode comment/uncomment function is useless and no comment highlight either. -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
Re: Bye from forks sends signals?
Hi Dmitry,
> Is this newsletter alive?:)
Yes, just not so noisy :)
> I'm playing around with `native` and ZeroMQ and found a curious behaviour.
>
> (unless (fork)
> (wait 2000)
> (bye))
>
> (setq Context (native "libzmq.so" "zmq_ctx_new" 'P))
> (setq ZMQ_REP 4)
> (setq Socket (native "libzmq.so" "zmq_socket" 'P Context ZMQ_REP))
> (native "libzmq.so" "zmq_bind" 'I Socket "tcp://*:")
>
> (buf
> Buffer 10
> (prinl "Waiting for messages")
> (when (= -1 (native "libzmq.so" "zmq_recv" 'I Socket Buffer 10 0))
>(prinl (pack "Error: " (errno)
This looks good.
> The `errno` is 4 (which is signal interrupt as i understand).
Yes. EINTR is 4 on most systems. You can see it with
: (sysdefs "errno")
-> EACCES
: EINTR
-> 4
or (vi "@lib/sysdefs").
> My assumption here is that `bye` throws some signal? Why else would it affect
> zeromq in the parent process?
> Actually, while writing this I found out about SIGCHLD which is apparently
> sent to parent on child's exit so I guess zmq_recv gets interrupted by that
> for some reason?
Absolutely correct. The child sends a SIGCHLD signal, which must be handled or
ignored.
I don't have libzmq at the moment and cannot test it, but I think it should be
something like
(while (lt0 (native "libzmq.so" "zmq_recv" 'I Socket Buffer 10 0))
(unless (== EINTR (errno))
(quit (errno) "Signal") ) )
> P.S. completely offtopic but since I'm here. I just noticed that semicolons
> aren't treated as comments. Why?
Comments in PicoLisp are # or #{...}#.
> Can it be enabled? Otherwise my Emacs' lisp-mode comment/uncomment function is
> useless and no comment highlight either.
There is no built-in way to change it. There are some Emacs libs for PicoLisp,
but I don't use Emacs and cannot be helpful here.
☺/ A!ex
--
UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
