Re: (no subject)

[Akim Demaille]

| > I'd prefer that you change XFile's open to handle this.  Would that be
| > a problem?
| 
| My feeble perl knowledge suggests this should do the trick:
| 
| sub open
| {
|   my ($fh) = shift;
|   my ($file) = @_;
| 
|   # WARNING: Gross hack: $FH is a typeglob: use its hash slot to store
|   # the `name' of the file we are opening.  See the example with
|   # io_socket_timeout in IO::Socket for more, and read Graham's
|   # comment in IO::Handle.
|   ${*$fh}{'autom4te_xfile_file'} = "$file";
| 
|   if (!$fh->SUPER::open (@_))
|     {
|       my $me = basename ($0);
|       my $file = ${*$fh}{'autom4te_xfile_file'};
|       croak "$me: cannot open $file: $!\n";
|     }
|   binmode $fh->SUPER
|     if ${*$fh}{'autom4te_xfile_file'} =~ /^\s*>\s*[^-]/;
| }
| 
| The "\s*[^-]" is there to ensure we don't put stdout in binary mode.

Is it really wrong anyway?

| (Note: is $file invalidated after the SUPER::open? If not, then why is
| ${*$fh}{'autom4te_xfile_file'} used in the error handler?)

I don't remember what motivated that, feel free to make it simpler.
Maybe it's simply the result of moving things around.