Paolo Carlini wrote: > ... it also fixes most of > the spurious regressions that happen if we artificially give a non-zero > capacity to a default-constructed string Correction: not most, *all* of them! :) :) Paolo.