I was putting an error message wrapper around some file operations like rename and link to get the error handling 'out of band' -- not have it inline w/the code so it obscured the flow of the code. Created a small wrapper:

  sub fileop($;$) {
    my ($op, @argv)=@{my $_=shift};
    use Carp qw(cluck);
    no strict 'refs';
    &{"CORE::".$op}(@argv) || do {
      cluck (P "ERROR: $! doing %s(%s): (%s)%s", $op, (join ', ', @arg
+v), $!,
        @_?" (extra info: ".$_[0].")":"");
      exit (0+$!)
    }
  }
# can call this with:
    fileop(['rename', $from_name, $tmpnam]);
    fileop(['link', $name_of_to_node, $from_name]);
[download]

Went to do the same with "unlink":

 sub sched_unlink($$) { fileop(['unlink', $_[0]], "from inode ".$_[1])
+ }
[download]

But had this coming out from perl:
&CORE::unlink cannot be called directly at <file> line <#>

Note, I had earlier, tried making such calls without the "CORE::" in front of them, but got errors about no such local routines being defined...so thought 'CORE::' was the way to go for perl built-ins...

Why the exceptions?... and do I just have to special case all the ones that give errors as they come up?

I didn't get from my last read of the docs that "unlink" was somehow radically different from "link"... Ideas?