On Wed, 15 Mar 2000, Jochen Stenzel wrote:
> > Are you talking about streams? I_SETSIG causing SIGPOLL?
>
> Ah! Somebody who has probably heard this as well.
>
> > ... You could actually say "whenever data comes
> > in through the stream head attached to file descriptor blah (ioctl
> > I_SETSIG), cause the kernel to SIGPOLL me with the details.
>
> This sounds very (interesting and) similar to the things the person told me. Is this
>designed for writing, too? Is it described anywhere? Do you remember the name of the
>book?
>
> You mentioned Solaris, so I suppose it's built in there ... ? If so, the question is
>if syscalls like select() already benefit from this or if it has to be used a special
>way.
>
> By the way, this could possibly be a solution of the EOF problem, couldn't it?
>(Assumed the kernel sends no message about EOF as "new data".)
>
> Thank you!
>
> Jochen
>
All this is described in Stevens Unix Networking volume 1, SIGPOLL and the
various POSIX asynchronous IO methods, it is nothing new but it is nothing
portable (yet).
A fd to a file is always readable under unix, therefore you should not
watch files using select as select will always return readable, have a
timer loop and poll is the only way, and EOF is an error on a socket, not
an error on a file.
Artur
