From: ahoward <ahoward@...>
Date: 2003-06-17T09:59:45+09:00
Subject: Re: popen, rsh, and File::NONBLOCK

On Tue, 17 Jun 2003 nobu.nokada@softhome.net wrote:

> io/wait in rough provides IO#ready?.
>
> $ ruby -rio/wait -e 'cat = IO.popen "cat", "r+";cat.puts "foobar";sleep 0.1;p cat.ready?;p cat.gets;p cat.ready?'
> 7
> "foobar\n"
> nil

> Also nonblock.rb is in io/wait.
>
> $ ruby -rio/nonblock -e 'cat = IO.popen "cat", "r+";cat.nonblock=true;cat.puts "foobar";p cat.gets;p cat.read(1)'
> "foobar\n"
> -e:1:in `read': Resource temporarily unavailable (Errno::EAGAIN)
> 	from -e:1

very cool - these will be in 1.8?

> Note that IO#gets waits till a newline even in non-blocking mode.

i did overlook that...

> Only access mode has meanings in the second argument to IO.popen.  NONBLOCK
> is ignored here.

i figured as much

> # NOCTTY affects to only terminal device.

if you are spawning an rsh command it is required though...

> Basically, pipe(2) doesn't take optional flags.  IO.popen just follows it.
> It would be possible to treat the flag as special...

i think it makes sense - but that's my opinion...  of course there would be
issues with fifos then where NONBLOCK *does* mean something...

IMHO it would be very useful to be able to do something like

  stdin, stdout, stderr = Process.new 'external.sh'

  stdin.puts command
  puts stdout.gets # should not block
  puts stderr.gets # should not block


essentially a means of sending commands to a process, getting back the stdout
and stderr, and being able to do this without blocking on io.  is this
sufficiently common to put together an RCR?

-a
--
  ====================================
  | Ara Howard
  | NOAA Forecast Systems Laboratory
  | Information and Technology Services
  | Data Systems Group
  | R/FST 325 Broadway
  | Boulder, CO 80305-3328
  | Email: ara.t.howard@noaa.gov
  | Phone:  303-497-7238
  | Fax:    303-497-7259
  | ~ > ruby -e 'p(%.\x2d\x29..intern)'
  ====================================