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

What you are missing is:

"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

Of note -- if STDOUT was redirected to a file, then trying to determine its size woudn't return valid values (presumably the ioctl would return an error).

So how would you determine the rows/columns of whatever STDOUT is pointing to (without opening another character device like "/dev/tty")?

Even if you aren't on linux, presumably there's a way to do it on your platform, no? I.e. if you have a 'tty' window, how would you determine the size of the window addressed by 'STDOUT'?

See my answer to previous poster -- that's a red herring as well as the unpack format.

Same guy, and no, the red herring is you're confusing c function ioctl with perl function ioctl

What you are missing is:

No, I didn't miss it. If you redirect STDIN/STDOUT/STDERR, then you need to give ioctl a handle which is connected to tty, so you open your own to /dev/tty, and it works regardless of the state of STDIN/STDOUT/STDERR

Even if you aren't on linux, presumably there's a way to do it on your platform, no? I.e. if you have a 'tty' window, how would you determine the size of the window addressed by 'STDOUT'?

I would use an abstraction like Term::Size::Any, I wouldn't mess around with ioctl

But in all honesty, unless I'm writing some kind of curses program, I would never need to know the size of the terminal

Confusing docs for system ioctl & perl ioctl -- got it.

So my error check has a high level of bogosity! Details! ;-)

Regardless of the bogus error check, it still tries to extract the row,col meaning that if ioctl(STDOUT...) is valid, it should display the rows & columns.

I understand it won't work if STDOUT is redirected, but neither will opening "/dev/tty", if the program isn't running with a controlling terminal but DOES have STDOUT directed to a tty. I.e. one can always design a failure case -- but I'm more interested in why what should be a 'success' case doesn't work, namely using ioctl with 'STDOUT' when STDOUT is attached to a terminal.

Here's a simple program that should work but doesn't. Why doesn't it (I believe I've addressed the error case you raised, as well as the unpack format)?

#!/usr/bin/perl -w
use strict;
use feature ':5.10';

{ no warnings "all";
  $SIG{__WARN__} = sub { };    #pretend "no warnings" works
  require "sys/ioctl.ph";
}

sub getwinsize (;$) {
  my $recheck=$_[0];
  my $winsize=0;
  state ($maxcols, $maxrows);

  return ($maxcols,$maxrows) if $maxcols && $maxrows && !$recheck;

  my $err = ioctl STDOUT, &TIOCGWINSZ, $winsize;
 
  unless ($err) {  
    print STDERR "ERROR: ioctl on STDOUT: $!\n";
    $err = ioctl fileno(STDOUT), &TIOCGWINSZ, $winsize;
    unless ($err) {
      print STDERR "ERROR: ioctl on fileno(STDOUT): $!\n";
    } else {
      printf STDERR "ioctl on fileno(STDOUT) worked\n";
    }
  } else {
    printf STDERR "ioctl on STDOUT worked\n";
  }

  my ($rows, $cols, $unused1, $unused2) = unpack "S!4", $winsize;

  printf STDERR "(rows=%u, cols=%u)\n", $rows//0, $cols//0;

  ($maxcols, $maxrows) = ($cols || -1, $rows || -1);
}

my ($rows,$cols) = getwinsize;

print "rows=$rows, cols=$cols\n";

# vim: ts=2: sw=2
[download]

My output:

ERROR: ioctl on STDOUT: Bad address
ERROR: ioctl on fileno(STDOUT): Bad file descriptor
(rows=0, cols=0)
rows=-1, cols=-1
[download]