[orbitcpp-list] back with a vengeance :)

[Andreas K. <ak...@ix...>]

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
:-)

(actually, it's already over since last thursday, but i just needed some
time for recovery from all those parties and time to visit my parents
and time to clean the flat and time to... oh well....;)

back to some "real work" then... 

* do you see any problems with having the skels use virtual inheritance?

* 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...

* some things concerning tonight's commit:

- it is not perfect. the string test works, the basic test works almost,
boolean fails to compile, i haven't yet tried "everything". i didn't
want to wait till i catch all those rather small bugs to get this commit
done because it has some effect on all of the compiler.

- 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.

- for the rest of the changes, see the cvs log entry.

cya
andy