[Ktechlab-devel] Please test before commit

[P Zoltan]

On Mon, 12 Oct 2009 23:56:07 +0200, Alan Grimes <agrimes at speakeasy.net>  
wrote:

>
> I was a bit panicked when I read that subject line given the number of
> bad commits I've made. =0, But thankfully, the latest revision I worked
> on was 503. =P I can move resistors around in 503 just fine.
>
> Wait a minute, my version does have the version of node.h reverted by  
> 506.
>
> hmm... =\
>
> Maybe if you do a clean-build it either does or doesn't work (not sure
> which), based on whether the MOC file is regenerated. I'm not sure which
> version of the MOC I'm running, so maybe I'll see it break on my end if
> I do a clean build.
>
> That said, my local version has many meaningless and or experimental
> changes that I have not committed. It is quite possible that among these
> are some that are actually important. =\
>

  Look in connector.cpp, where the moveTo slot is connected to the moved  
(?) signal. There the signature of that slot was still  
moveTo(double,double). So it would have worked correctly if on _all_ the  
places the same moveTo(const double, const double) would have been used.  
Maybe I should have changed the connnector class, and a few other places...

> In general, "const" is annoying as hell to type  and read all over the
> place again and again. BUT, it does allow the compiler to help find
> errors in your code, confirm design assumptions, and even allow the
> compiler to find more optimal instruction orderings.


  Yes, agreed.

>
>>   It would be nice if we wouldn't introduce regressions; it something
>> doesn't "want" work, better talk about it on the list.
>
> Sometimes it is unavoidable. Making the linear algebra engine more
> correct, basically broke many circuits which were relying on the old
> engine to fudge things just enough to work. However, no valid simulation
> is possible unless the linear algebra system is correct.
>
> Also, improving the models for nonlinear components seems to involve
> making sure that the physical constants used are correct to the limits
> of science. (see physics.h), and to make sure the equations are well
> implemented. Unfortunately, I was not able to make it work. =(

  I want to have a look at the equation solver, after cleaned up the  
document's destructors (so they won't crash...).