Date: Sat, 6 Apr 1996 17:16:20 +1000 From: Bruce Evans <bde@zeta.org.au> To: hackers@FreeBSD.org, jkh@time.cdrom.com Subject: Re: Any clues as to why this fails? Message-ID: <199604060716.RAA14868@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>#include <signal.h>
>#include <fcntl.h>
>void
>handler(int sig)
>{
> printf("Signal %d received\n", sig);
>}
>main()
>{
> int i, fd;
> signal(SIGIO, handler);
> fd = open("/dev/cuaa1", O_RDONLY | O_NONBLOCK | O_EXCL);
> printf("%d\n", fd);
> i = fcntl(fd, F_GETFL, 0);
> printf("%d\n", i);
> printf("%d\n", fcntl(fd, F_SETFL, i | O_ASYNC));
> printf("%d\n", fcntl(fd, F_SETOWN, getpid())); /* This call returns -1 */
> pause();
>}
>I've read the man page for fcntl() and it really appears as though
>that F_SETOWN call should work. Any clues as to why not? This
F_SETOWN on a tty only works for controlling terminals. This is because
F_SETOWN is implemented as tcsetpgrp() and tcsetpgrp() is specified by
POSIX. To work like you want, F_SETOWN would need a separate pgrp
entries in the tty struct. Sockets work like you want because there is
no POSIX pgrp entry to conflict with. F_SETOWN doesn't work on anything
except ttys, sockets or perhaps pipes. To work as documented in fcntl.3,
F_SETOWN would need to have a pgrp entry in the filedesc struct. This
would probably be useful - it would allow sending SIGIO to arbitrary
sets of sufficiently privileged processes.
The test program has many bugs.
- <sys/types.h> isn't included.
- printf() gives undefined behaviour in signal handlers.
...
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604060716.RAA14868>