read/writes to pipes with O_NDELAY set

[Matthias Pfaller]

according to the manuals of our Unix-box reads and writes to a pipe
with O_NDELAY set should behave as follows:
>
> If the file being written is a pipe (or FIFO) and the
> O_NDELAY flag of the file flag word is set, then write to a
> full pipe (or FIFO) will return a count of 0. Otherwise
> (O_NDELAY clear), writes to a full pipe (or FIFO) will block
> until space becomes available.
>
>
> When attempting to read from an empty pipe (or FIFO):
>
> If O_NDELAY is set, the read will return a 0.
>
> If O_NDELAY is clear, the read will block until data is
> written to the file or the file is no longer open for
> writing.
>
Minix returns EAGAIN; what is correct, what says Posix?
Matthias Pfaller

[Doug Ingraham]

POSIX changed O_NDELAY to O_NONBLOCK. The reasoning is detailed in section
B.6 pp250-251. They say that "Systems may choose to implement both O_NDELAY
and O_NONBLOCK, and there is no conflict as long as an application does not
turn both flags on at the same time." I am slightly confused by this. I
think that the use of O_NDELAY results in a non-standard conforming program.

On the matter of blocked reads there is line on page 255 that I think will
answer your question.

"Where the standard requires -1 returned and errno set to [EAGAIN], most
historical implementations return zero (with the O_NDELAY flag set: that
flag is the historical predecessor of O_NONBLOCK, but is not itself in the
standard). The error indications in the standard were chosen so that an
application can distinguish these cases from end-of-file. While write()
cannot receive an indication of end-of-file, read() can, and the Working
group chose to make the two functions have similar return values."

On page 112 are the rules that define what must happen:

When attempting to read from an empty pipe (or FIFO):

1) If no process has the pipe open for writing, read() shall return zero
to indicate end-of-file.
2) If some process has the pipe open for writing and O_NONBLOCK is set,
read() shall return -1 and set errno to [EAGAIN].
3) If some process has the pipe open for writing and O_NONBLOCK is clear,
read() shall block until some data is written or the pipe is closed
by all processes that had opened the pipe for writing.

When attempting to read a file (other than a pipe or FIFO) that supports
non-blocking reads and has no data currently available:
1) If O_NONBLOCK is set, read() shall return -1 and set errno to [EAGAIN].
2) If O_NONBLOCK is clear, read() shall block until some data becomes
available.
3) The use of the O_NONBLOCK flag has no effect if there is some data
available.

I think that MINIX is correct in returning EAGAIN as far as POSIX is
concerned. I am not yet running 1.5 so I can't check to see if O_NONBLOCK
is also supported.

I think this answers your question.

--
Doug Ingraham (SysAdmin)
Lofty Pursuits (Public Access for Rapid City SD USA)
uunet!loft386!dpi