I am trying to figure out how perl determines the order in which to output STDOUT and STDERR text, both to a terminal and to a file. Consider the following code:

#!/tool/bin/perl -w
use strict;

my $i = 0;
my @array;


for ($i = 0; $i < 100; $i++)
{
    printf("element $i of array = ");
    printf("%s\n", $array[$i]);
}
[download]

If I run this code in a terminal, the output is as follows:

> ./stderr_test.pl
Use of uninitialized value in printf at ./stderr_test.pl line 11.
element 0 of array =
Use of uninitialized value in printf at ./stderr_test.pl line 11.
element 1 of array =
Use of uninitialized value in printf at ./stderr_test.pl line 11.
element 2 of array =
Use of uninitialized value in printf at ./stderr_test.pl line 11.
element 3 of array =
Use of uninitialized value in printf at ./stderr_test.pl line 11.
element 4 of array =
...
[download]

As you can see, the stdout text is interleaved with the stderr, which is what I would expect based on the order of the printf statements. If I redirect the text to a file however, the result is different:

> ./stderr_test.pl > & log
[download]

where log contains:

Use of uninitialized value in printf at ./stderr_test.pl line 13.
Use of uninitialized value in printf at ./stderr_test.pl line 13.
Use of uninitialized value in printf at ./stderr_test.pl line 13.
...
element 0 of array = 
element 1 of array = 
element 2 of array = 
...
[download]

As you can see, in this case perl first outputs all the stderr output, then outputs all the stdout output. Why does it do this?

Why is the behavior different when I output to a file vs. the terminal?

More importantly, how can I force the file output to appear in the correct order? Debugging is much easier when printed statements appear in programatic order.

I tried adding the following to my code above:

open(STDERR,'>&', STDOUT);
[download]

but that didn't make any difference in my output file.