That is close to a test I've used:

my $o= pack( 'I', 0x04030201 );
if(  substr($o,-4) eq "\04\03\02\01"  ) {
    return "big";
} elsif(  substr($o,0,4) eq "\01\02\03\04"  ) {
    return "little";
} else {
    die "Strange endian";
}
[download]

Is substr() offset always in bytes?

Note that the "I" format can result in more than 4 bytes on some platforms.

Yeah, I suppose you could use 'L' then.

I suppose you could figure out the size of integers with code like:

@b = unpack('C*', pack('I', 0));
return scalar(@b);
[download]

Then you could make the endian code more generic with:

@b = unpack('C*', pack('I', 0));
$sizeof_long = scalar(@b);

@c = (1..$sizeof_long);
$i = pack('I', hex('0x0'.join('0',reverse @c)));
$big = pack('C'.$sizeof_long, reverse @c);
$lit = pack('C'.$sizeof_long, @);

if ( substr($i, 0, $sizeof_long) eq $big ) {
  return 'big';
}
elsif ( substr($i, 0, $sizeof_long) eq $lit ) {
  return 'little';
}
else {
  return "strange";
}
[download]

Good point. I don't do anything with "wide" strings in Perl yet so I usually forget to take that into consideration. But here we are making the strings so we know they'll be byte strings and we can use what I wrote or: length pack "I", 0 to find the number of bytes in an IV.

I thought I had used a "native 32-bit integer" before for endianness testing, but for some reason couldn't think of "L". That certainly makes more sense than resorting to substr. (:

        - tye (but my friends call me "Tye")

    #!/usr/bin/perl -w

    use strict;
    use Config;


    if ($Config{byteorder}    == 4321)     {
        print "Big endian\n";
    }
    elsif ($Config{byteorder} == 1234)     {
        print "Little endian\n";
    }
    elsif ($Config{byteorder} == 87654321) {
        print "Cray\n";
    }
    elsif ($Config{byteorder} == 3412)     {
        print "Weird order\n";
    }
    else {
        # Not sure if it can get here
        print "Unknown byte order\n";
    }
[download]