William Coleda <[EMAIL PROTECTED]> wrote: [ note: slightly modifed version number ;) ]
> So Dan's string stuff has been rolled into cvs main (along with a huge > # of leo addon patches...).... Yeah. The string stuff has again reached the functionality it had earlier, e.g. printing utf8 strings works again - at least as far as the tests go. > What's the current status of the new GC system? Still need testing? The generational GC system isn't yet functional. To get it running it needs: 1) interpreter support with two opcodes like: - scope_enter I_need_timely_destruction - scope_exit i.e. hints to the GC system, if a new generation should be created for a scope or not. 2) a core change in register frame handling As outlined in an earlier thread, we currently can't do timely destruction at scope level, as Parrot registers keep even dead objects alive. Only when the register frame is discarded, these references are gone. This implies that a scope needs its own register frame just as currently a subroutine call creates a new register frame. > I'd recommend postponing any of the architectural/design items that > were slated for 0.1.12 to 0.2.0 at the latest, perhaps another point > release if we can get to them sooner... However, until we get Dan (#1) > back, there's not much we can do here, I think. Basically yes. There are still enough items to work on that don't need architectural changes, which the big cheese ought to approve. OTOH some changes should be done more sooner then later, either for reasons of correctness or just to improve or even allow progress in other parts of the interpreter. I'll post a list of proposed changes after 0.1.2 is out. As releasing 0.1.2 is already delayed, I'd like to shorten the usual procedure a bit like so: - feature freeze starts now Bug fixes, docu updates and such are still very welcome. This also includes the installation process but that's not a must. - no more checkins after Sun 6th, 12.00 GMT > Will Coleda (aka #6) leo
