Re: [orbitcpp-list] back with a vengeance :)
Status: Beta
Brought to you by:
philipd
|
From: Phil d. <ph...@us...> - 2000-03-30 11:06:50
|
Andreas Kloeckner wrote: > > Hi Phil & folks, > > i'm back from the dead, finally... the calculus exam is over! hooray! > two dumb weeks of studying, six hours of examination spread across two > days, phew... > (we'll see about the "hooray" when i get my results, but for now i'm > just happy that i will definitely not have to do all this stuff again > :-) > I know how you feel. The time after my finals was fantastic. The best thing is being able to watch tv without feeling guilty for not revising! > back to some "real work" then... > > * do you see any problems with having the skels use virtual inheritance? > Don't think so. The skeletons can be as elaborate as we like, since AFAIK there's no way we can make them 'binary compatible'. > * the 0.26 release is awfully broken. i noticed that i slipped in some > code that dumb-loops the compiler on any occurence of inheritance. we'd > better get 0.27 out the door before anyone notices... > hmmm... How about after the weekend? I should have the unbounded sequence stuff finished by then. > * some things concerning tonight's commit: > Doh! I've got some sequence code to commit. Better get merging... > - new: the notion of "type containers". up to now, the compiler just > threw away any types declared inside an interface. to remedy this, i've > introduced type container classes, which are classes that sit in the > same namespace hierarchy as their real-life stub counterparts, but under > "_orbitpp::type_container". the compiler translates iface-scope types > into types inside these containers, from where they are imported into > stubs and skels by means of inheritance. unfortunately, even though > these container classes don't have data members, gcc makes them have > size 1. thus, the stub binary footprint is no more the same as C > ORBit's. yet, i hope this does not > matter because gcc hopefully never accesses this memory. Hmmm sounds dodgy ;-) Could it be something to do with rtti? > - for the rest of the changes, see the cvs log entry. Got it. Cheers, Phil. |