> Mattia Barbon <[EMAIL PROTECTED]> wrote: > > > Hel= lo, > > this patch: > > 1) removes the necessity of having "#define enu= m_class_Foo -1" > > in the dynpmc file > > 2) makes DYNSUPER fail at= compile time instead of at runtime > > 3) makes MDD declarations work, = as long as there isn't a circular > > dependencies and PMCs are loade= d in the correct order > > > All tests pass (Linux, x86), but this pat= ch does not affect > > output for core PMCs anyway. > > s/not/almost = not/ :)
Oops. Correct. > The constness of the MMD_init temp tabl= e got dropped. We should keep > these tables const. (Same holds BTW for = the temp vtable templates). Why? In the current code they are create= d on stack. Does it really matter if they are const or not, in this case= ? They are copied anyway. > So what about: make these tables const (ag= ain) and patch in the > necessary dynamic changes during MMD registering= or after during vtable > copying respectively. I can do that by pa= ssing an array of integers to the registration function (for example). I= just can't see what is gained by this approach. Regards Mattia
