At 08:55 AM 7/5/2002 -0500, David M. Lloyd wrote: >On Wed, 3 Jul 2002, Melvin Smith wrote: > > > At 09:51 PM 7/3/2002 -0400, Josh Wilmes wrote: > > >I know there was some talk about this extra "address" parameter recently, > > >but i'm not sure what the upshot of it is. Right now, tcc is complaining > > >loudly because the init functions for parrotsub and parrotcoroutine don't > > >match the init_method_t type in the _vtable structure. > > > > > >What's the deal here? > > > > I made the PMC init method take an int argument sometime around > > January, and at the time we agreed it would be useful to have polymorphic > > constructors. Then it sort of silently got removed. > > > > I suppose the patch was done around the same time I commited > > the sub/coroutine PMCs, which would explain why they got out of sync. > > > > Passing an optional arg to the constructor can possibly save us cycles > > in the long run, for examples, the PMCs that need to know how to size > > themselves. Unless we do lazy allocation, aggregates usually preallocate > > at construction time, then resize later. For languages other than Perl > > that might need a "struct" type, it makes sense to request the correct > > allocation up front. The size could be bytes, elements, or kilometers. > > It doesn't need to be just size, as in subs/coroutines. > > > > In any case, it wasn't doing any harm, so I'm not sure why it got > > removed. > >It got removed because it wasn't in the spec... Dan directed that we >replace it with a version of init that accepts a PMC argument >(init_pmc_method_t) that can be used to send in initial size or whatever >else you can dream up.
Ah yes, sorry I wasn't poking at anyone. I just remember being distracted and then, *poof*, it was gone. I recall the (much needed) monster patch for PMC cleanup. -Melvin
