Mailing List Archive

[perl #112164] stat() warns about unopened filehandle on tied handles
On Sat Mar 31 10:55:15 2012, wrote:
> On Sat, Mar 31, 2012 at 7:39 PM, Father Chrysostomos via RT
> <> wrote:
> > fileno() there is not a real fileno call, but tied(*FOO)->FILENO.
> Sure. That was just to confirm the file was indeed opened.
> >> Tied filehandles are still incomplete. �sysopen(), truncate(),
> >> flock(), fcntl(), stat() and -X can't currently be trapped.
> >
> > So it sounds as though we would need to implement STAT on tied
> > filehandles before being able to fix this properly (i.e., for all
> > handle-tie classes).
> I think that's a good long term goal,

I’d like to get that done for 5.18.

What should the interface be for the STAT call?

Should it be expected to return a 13-element list? Should those
returned values be associated with the _ handle?

What should happen if it returns a list with fewer elements?

Should pp_stat croak, or assume (@ret, (undef)x(13-@ret))?

If it croaks, should it be forgiving with ‘return undef’? (Probably not.)

I think I’m in favour of having it croak if @ret != 13 && @ret != 0.
And the values should be saved for _.

Filetest calls like -r and -w can be implemented in terms of STAT, but
what do we do about -t -T -B?

> but a more immediate fix is that
> stat() should return the empty list for tied filehandles. That better
> reflects that stat() is failing (because handles don't support STAT),
> and avoids giving misleading information (e.g. "unopened") about the
> handle.

This can be the fallback for handles without STAT.


Father Chrysostomos

via perlbug: queue: perl5 status: open