I wrote: >sprintf. The common underlying function is do_sprintf(). But it turns out that CORE::sprintf has a $@ prototype, and so doesn't invoke this bug after all. I reckon the underlying bug is CORE::printf having been given the prototypish behaviour of CORE::print rather than that of CORE::sprintf. This accounts for both this stack bug and the undocumented use of $_. -zeframThread Previous | Thread Next