See also Pack/Unpack Tutorial (aka How the System Stores Data), s!* looks suspicious, what I'd usually do is try every option :)

update: Oh look, this is a perlfaq8 item

How do I get the screen size?

If you have Term::ReadKey module installed from CPAN, you can use it to fetch the width and height in characters and in pixels:

    use Term::ReadKey;
    ($wchar, $hchar, $wpixels, $hpixels) = GetTerminalSize();
[download]

This is more portable than the raw ioctl, but not as illustrative:

    require 'sys/ioctl.ph';
    die "no TIOCGWINSZ " unless defined &TIOCGWINSZ;
    open(TTY, "+</dev/tty")                     or die "No tty: $!";
    unless (ioctl(TTY, &TIOCGWINSZ, $winsize='')) {
            die sprintf "$0: ioctl TIOCGWINSZ (%08x: $!)\n", &TIOCGWIN
+SZ;
    }
    ($row, $col, $xpixel, $ypixel) = unpack('S4', $winsize);
    print "(row,col) = ($row,$col)";
    print "  (xpixel,ypixel) = ($xpixel,$ypixel)" if $xpixel || $ypixe
+l;
    print "\n";
[download]

So the answer here is that I have to open "/dev/tty" each time I call it rather than reference my already-open "STDOUT"?

Hmmm....I'm disappointed that I need to open another file handle and can't get the rows & columns associated with my already open FH, "STDOUT", directly...

As for using 'Term::ReadKey', I thought of using something along those lines (or a Curses func), but wanted something that was minimal overhead, as this was called every loop, and portability isn't a major consideration, considering the nature of the program this is being used in is showing linux process and mounted-dev I/O.

If I look at /proc/$$/fd/1, I see a softlink to /dev/pts/1 (i.e. a character, tty type device). Indeed, if I use your 2nd prog and substitute in "/dev/pts/1" instead of "/dev/tty", I get the same result (ie. it works).

Shouldn't STDIN, STDOUT map to fd[0], fd1 => /dev/pts/1 on some level?

I.e. Shouldn't I be able to use some mapping function on STDOUT to get a file descriptor that's suitable for ioctl?

So the answer here is that I have to open "/dev/tty" each time I call it rather than reference my already-open "STDOUT"?

I don't have linux, but no, I don't think that is the answer.

I think part of the answer is to do error checking correctly like the faq entry

I also think using S4 is part of the answer, not being an expert on pack/unpack and not having a /dev/tty, I can't run any tests.

Hmmm....I'm disappointed that I need to open another file handle and can't get the rows & columns associated with my already open FH, "STDOUT", directly...

Sounds premature if you ask me, avoiding abstractions always involves some growing pains :)

Shouldn't STDIN, STDOUT map to fd[0], fd[1] => /dev/pts/1 on some level?

They do

$ perl -le " print fileno($_) for STDIN, STDOUT, STDERR "
0
1
2
[download]

See, fd 0,1,2. Now regarding tty, using operator -t to test if filehandle is opened to a tty, you can see 0,1,2 are connected to a tty

$ perl -le " print 0+-t $_  for STDIN, STDOUT, STDERR "
1
1
1
[download]

but not when you do redirection, here is STDIN not connected to a tty

$ perl -le " print 0+-t $_  for STDIN, STDOUT, STDERR "  < NUL
0
1
1
[download]

here is STDIN and STDERR not connected to a tty

$ perl -le " print 0+-t $_  for STDIN, STDOUT, STDERR "  < NUL  2> NUL
0
1
0
[download]

here is STDOUT not connected to a tty

$ perl -le " print 0+-t $_  for STDIN, STDOUT, STDERR " > out.txt

$ cat out.txt
1
0
1
[download]

I.e. Shouldn't I be able to use some mapping function on STDOUT to get a file descriptor that's suitable for ioctl?

ioctl says you don't need to, the faq item doesn't, and neither does Term::Size::Any

See ioctl, if the return value is true, its not an error.

My manpage says:

RETURN VALUE
       Usually, on success zero is returned.  A few ioctl() requests  
+use  the
       return  value as an output parameter and return a non-negative 
+value on
       success.  On error, -1 is returned, and errno is set appropriat
+ely.
[download]

Note the phrase: "Usually, on success, zero is returned."... Some return a value as an output param, but this call isn't one of them. My error check code appears to be valid according to the above, though I noticed an oddity if you print the value returning from ioctl in this code:

 require 'sys/ioctl.ph';
    die "no TIOCGWINSZ " unless defined &TIOCGWINSZ;
    open(TTY, "+</dev/tty")                     or die "No tty: $!";
    my $ret=ioctl(TTY, &TIOCGWINSZ, $winsize='');
    printf "ret num=%d, retstring=%s\n",$ret,$ret;

    unless ($ret) {
            die sprintf "$0: ioctl TIOCGWINSZ (%08x: $!)\n", &TIOCGWIN
+SZ;
    }
    ($row, $col, $xpixel, $ypixel) = unpack('S4', $winsize);
    print "(row,col) = ($row,$col)";
    print "  (xpixel,ypixel) = ($xpixel,$ypixel)" if $xpixel || $ypixe
+l;

    print "\n";
[download]

The output is:

ret num=0, retstring=0 but true
(row,col) = (79,80)
[download]

That is bizarre! 0 = true?!

In any event, that's a red herring, as my code works when I use "/dev/tty" as a device instead of STDOUT. Same goes for the unpack format.

I used S!*, which is likely "more correct" than S4 shown above.

The "!" after the "S" says to use the platform's native size for a short instead of fixed a fixed 16-bit size. Since the struct is declared as a short on my platform, I figured it was best to use my platform's native "short" size rather than a fixed 16 bit value -- wouldn't you agree??

The "*" vs. "4", makes no difference in this case as there are only 4 values to receive the output. "*" tells unpack to unpack as many as are in the struct -- so if the struct was 6 shorts long instead of 4, it would return 6 values, the last 2 of which would be thrown away as I only have an array of 4 vars to receive it. It would be like doing:

my @ar=(1,2,3,4,5,6);
my ($x,$y,$z)=@ar;
[download]

$x-$z get values 1 through 3, and values 4,5,6 get ignored.

All of the above covers the stuff that "doesn't make a difference". Then comes the part that doesn't make sense.

As you note, fileno() on STD(IN,OUT,ERR) all return 0,1,2 as expected. Then why can't I use ioctl(STDOUT) with ioctl and have it work?

My manpage says:

Manpage? I already linked to ioctl before, perldoc -f ioctl