defunct sh processes

[Aharon Robbins]

Greetings. Re this:

> Date: Mon, 17 Nov 2003 01:23:10 +0200 (EET)
> From: Juha Markkula <address@hidden>
> To: address@hidden
>
> Hello
>
> If you run for example this command in bash:
>
> for ((i=0;i<100;i++));do echo $i;done|gawk '{ ("echo " $1 )|getline \
> foo;system("sleep 1");print foo }'

The ``bug'' is in your awk program.  If you put

        close("echo " $1)

into your program after the print, things will work fine.  Gawk
can't tell if a subchild is going to continue to provide more input
or not, you can legitimately use

        while ((command | getline) > 0)
                ....

I suppose I could make the code a little smarter about processes that
have died, but only at the cost of increasing its complexity.  By
adding a call to close, your program will be correct and will work
without problems.

I'm very busy right now, but I'll try to put looking at this on my
TODO list.

> In one of my programs I have to use awk's "command"|getline -construct in
> this way and fork() stops working for gawk after a while because process
> table fills up. My solution has been using mawk which does clear it's
> returned child processes (wait() for them perhaps) at least at some time
> intervals. If I run the example above using mawk it does not have more
> than one defunct process at a time.

I've never claimed to be smarter than Michael.  If mawk works for you,
go for it.

Arnold