jpayne%bbn-vax@sri-unix.UUCP (03/08/84)
From: Jonathan Payne <jpayne@bbn-vax> It sure would be neat if EOF's could be sent down pipes. I've heard a rumor that you can do that in 4.2. Is that true? Something like write(fd, nul, 0); or something like that anyway. Is it actually hard to implement in any version of UNIX? Doesn't seem so. Wouldn't be very compatible with the rest of the world ...
dedwards%usc-isi@sri-unix.UUCP (03/09/84)
From: Howard S. Weiss <dedwards@usc-isi>
It was done years ago on V6 at RAND - in fact here is a manual page:
EOFP(II) 8-March-1977 EOFP(II)
NAME
eofp - write an end-of-file on a pipe
SYNOPSIS
eofp(fd);
DESCRIPTION
The pipe I/O routines have been modified to allow single
end-of-files to be transmitted. It works as follows: a call
to eofp guarantees that, after the real data is read, the
next read on the pipe will return with 0 characters. Further
writes may be done after writing an EOF, but they will be
suspended until the EOF actually gets read.
SEE ALSO
pipe (II)
BUGS
Writing 0 characters on a pipe now has very special
meaning--one's code should be checked to see that it disal-
lows all writes of 0 characters unless really desired.
-------boyd@basser.SUN (Boyd Roberts) (03/18/84)
> It sure would be neat if EOF's could be sent down pipes. I'm sure that what you need here is not EOF's down pipes but a *protocol*. After all, EOF means that the other end of the pipe has been closed, and not some other magical event. > Is it actually hard to implement in any version of UNIX? I'm also sure that it wouldn't be very hard to break pipes too. At least it'd be a step forward in the apparent quest to break everything, that up till now worked in the stated way. > Wouldn't be very compatible with the rest of the world ... Exactly, it wouldn't be very compatible. But I see that it's already been done at RAND with that ridiculous eofp() junk. I just dread the thought of "eofp() meets _filbuf()".
obrien%rand-unix@sri-unix.UUCP (03/20/84)
From: Michael_OBrien <obrien@rand-unix> "eofp()" solved a serious problem we had with pipes, which we were forced to use because of PDP-11 address space restrictions. It actually worked quite well. However, I should point out that it existed only in our highly hacked V6 kernel (I spent most of my first couple of years here unhacking that kernel and throwing stuff away). We no longer support, use, or recommend "eofp()" and it doesn't occur in any of our kernels any more.