$ perl -MPOSIX -wle'print fpathconf(*STDOUT, _PC_PIPE_BUF)'
Argument "*main::STDOUT" isn't numeric in subroutine entry at -e line 
+1.
4096

$ perl -MPOSIX -wle'print fpathconf(*FOO, _PC_PIPE_BUF)'
Name "main::FOO" used only once: possible typo at -e line 1.
Argument "*main::FOO" isn't numeric in subroutine entry at -e line 1.
4096
[download]

And yes, they are being interpreted as zero.

$ perl -wle'print 0+*STDOUT'
Argument "*main::STDOUT" isn't numeric in addition (+) at -e line 1.
0

$ perl -wle'print 0+*FOO'
Name "main::FOO" used only once: possible typo at -e line 1.
Argument "*main::FOO" isn't numeric in addition (+) at -e line 1.
0
[download]

His original code gives no warnings. And 0 + $wd doesn't evaluate to 0. I think it evaluates to the address of the object.

Looking at the code for fpathconf in glibc, I see that, for 16 of the 21 configuration parameters, fd is ignored—except to check that it is >= 0. So I think what's happening is that when the objects are being allocated above (void *)((unsigned int)-1 >> 1), when it's passed as an int, it's negative, so it gives an error.

Ah! He said glob in Re^2, but he's actually passing a ref to a glob in the OP (which becomes a number with no warning).