Re: [Simulavr-devel] repositories

[Onno Kortmann]

Hi again,
> >And yes, to backport this to CVS would be really a lot of work. It's
> >possible, but there are really much changes and the question is: take it
> >all without review and trust, that there are not really big bugs in it
> >or try to cherrypick the changes step by step?
>
> Trust, but verify?
Well, I try not step on anyone's feed here, BUT I think that the current state 
of simulavrXX is very much 'development'. Without being too deep into the 
code, I found several memory leaks, some other bugs and missing code (e.g. 
the ICP for timer1) and fixed several of these issues. Thomas recently did a 
lot of additional work - and I didn't even had the time yet to really look at 
it, such as fixes in my tracing code (thank you!) and further enhancements. 
Ask him for the very latest state of the project :-)

I feel that a 'keep it stable attitude' is a good idea here, but to keep it 
stable, it has to be stable in the first place.
Isn't that the distinction between stable and dev.? In dev. one is free to 
move things around and occasionally break something while improving the 
overall structure or to make room for future improvements; for stable, only 
important fixes are made. 

That said, I am of course open to suggestions regarding my changes :-)

> Probably the thing to do is to convert to git by copying a git repository.
> Do enough tests to make sure that what used to work still works.
That should be a TODO item: Simulavrxx lacks good tests IMO. On the other 
hand, some of my restructuring of the code would make a lot of class-based 
regression tests superfluous if they cover the current CVS code base. I 
simply think that there is still a lot of possibilities to improve the core 
structure of the simulator and I think that detailed regression test suites 
should be written as soon as there is a stable, 'clean' and 'good' 
architecture for simulavrxx - but not before. What is applicable should be 
used to test all changes, though - and I think this would be another TODO 
item, to get the current regression test suite to work cleanly :-)

So maybe it is a good idea to assemble a list of TODO items which *could* be 
done as improvements to simulavrxx and then go from this list to find a list 
of changes (through discussion here) which *should* be done before a stable 
release is created? If wanted, I could write down (into TODO) all the ideas I 
have about simulavrxx. - ?

Regarding the 'template stuff' (from your other post): I thought that it was 
giving the most abstract view and the most reduction in code size and I also 
have the opinion that in 2009, one ought to use templates if it makes sense. 
Compilers should support them. I am somewhat flexible in my opinion here, but 
I think that a possible replacement of the template stuff should be at least 
as concise and clear as mine is IMHO :-)
And it replaces A LOT of redundant code and creates a central point to attach 
tracing probes to the I/O parts of the simulator.


Best regards,

Onno