Re: [Simulavr-devel] test mechanism

[Thomas Klepp]

Hi Michael,

now I have time to answer.

> I thought it was something like that.
> I'd inferred that methods that referred to
> "non-existent" methods would need similar treatment.

This problem, that SWIG dosn't "translate" some operator methods comes 
from, how SWIG creates interfaces. All class methods, which SWIG found 
will be encapsulated in just normal C functions. And, for example for 
python, this functions will be collected in tables, which represent on 
python side for example a class. That SWIG now can't "translate" some 
methods is because this mechanism. A translation function itself can 
handle operator methods like a assignment. So it's no danger for 
malfunctions because of not translated operator methods. It's only, that 
such operator functions couldn't translated.

> In any case, the more general questions are what is to be development
> protocol and will it change while I am working on something messy.

Yes, that's really a question. But I can't answer this. Nobody from 
sinulavr maintainers has given a statement, what's the way to future. 
Stay on CVS, maybe backport changes from git (but this is now a lot of 
changes) or switch to git or whatelse.

> > My question points to that, that in git repo we use a derived class from 
> > RWMemoryMember, called IOReg. Your code would work too for registers, 
>
> This is different from the CVS version?

Yes, IOReg gives a connection between hardware units and rw memory 
array. Before you had external RWMemoryMember classes, which connect 
memory array to hardware unit. Now IOReg registers are member of the 
hardware units.

Before it was (for example):

rw[0x63] = new RWPin(this, portg);

Now it's coded as:

rw[0x63] = &portg->pin_reg;

I think, this idea from Onno is easier to understand and easier for 
changes and maintainance.

> I made sbi and cbi virtual so they could be overridden.
>
> Whose git repo is "git repo"?

Yep, also a good question. :-) My repo is a fork of Onno's, but we have 
agreed, that's not a good idea to have more than one. The most of 
reasons are connected, how this forks can be synchronized, the best is, 
to get it fully synchronized. But for that it's not necessary to have 
forks. But Onno is busy with other things and so I have made some 
development on my fork. The plan is, if Onno have time, to sync my fork 
to Onno's repo and then to delete my fork. In the moment, my master 
branch is younger (and more current) than Onno's.

> I expect to need to use git eventually,
> so what is the best way to make progress?

The easiest starting point for you (in my opinion) would be: install 
git, pull (for example) my master branch with "git clone" (see "git help 
clone" for help). So you get a workarea and a view what's happen here. 
bootstrap, configure and make work as usual (for you it's necessary to 
give configure also the option --enable-python to get the python 
interface too) I have tried to describe something, how to use git, see 
the wiki on my repo site.

For your own changes it's the best to make your own development branch, 
starting from master branch. So (and if you don't push this branch to an 
remote repo) you get a private branch for your changes. Later it's 
possible, to merge back your chnages to master branch or to pack a 
patchbomb of your changes to send it by mail, to sync new changes on 
master (after "git pull master") to your private branch and so one. And 
of course, you can delete also branches, if you want. :-)

And because git don't need a server, you can try all operations before 
on local clones of a repo. (just to try or to learn more about git, I 
have made it also in this way).

cu, Thomas