1.) "unit foo(bar1,bar2,...);" : You are absolutely right to let the compiler call the inits of the parental units implicitely.
2.) "inherited <someproc>;": I confused it with FPK-Pascal, which already has funcoverloading, I think. So it may be implemented after function overloading is implemented. Till this moment one might use
unit communication(graph,keyboard,mouse); ...blablabla procedure mouse_reset; begin mouse.mouse_reset; (*instead of "inherited mouse_reset;"*) ...logging end;
3.) virtual procedures (the circle() in Peters mail) I think its to much work to program right now. Lets first do it only in one direction. For now inherited procedures should be possible and enought.
4.) libraries Why make a difference between static and dynamic libraries? This difference is fundamental only on DOS/WIN-based machines. The handling is mostly the same (except -fPIC and -static) AFAIK on Linux. The best IMHO would be to make a library like that:
library test(testsensor,testmachine,testuser); (*$static*) or (*$dynamic*) ...blablabla which produces test.a(or test.so/.dll) from testsensor.o,testmachine.o and testuser.o and test.gpi from test*.gpi.
But I think that should be postponed for 2.2
A last question: What is the schema of the automatic type-conversion? Is there a global schema for converting between ordinal,real and string types or is it only done from case to case? I guess there is a mail somewhere in the archive.
Hans