> Is this documented ? I mean: will all implementations return > -1 in the error case ? If it isnt, it should be. Im +1 on this, assuming there is a _reason_ for it - ie, that some profiling shows this a hot-spot improved with the patch. If is just a "for-the-sake-of-it", Im -1 simply due to the KISS principal. If-it-aint-broke-dont-fix-it-ly, Mark.