[ODE] Making it build again

[John Miles]

There used to be a lot of warnings when building with dReal=float under VC6,
but all but one or two of them were fixed several few months ago.  I don't
remember who fixed them.

Best practice dictates that the code compiles without warnings.  In MSVC, I
usually promote warnings to errors and build at the most-sensitive warning
level (/W4).  I do disable a few particularly-stupid messages that should
never have been warnings in the first place ("Unreferenced formal
parameter," "Conditional expression is constant," and the like).

-- jm

>
> > > If you're not treating warnings as errors, well, you should! (I
> > > don't want to argue about it, but you can't find anyone who's any
> > > good who wouldn't agree).
>
> > I was never arguing that we should ignore warnings, but it's a matter
> > of priority.  If I have broken ODE so it doesn't compile, I'll fix it
> > ASAP.  If I've introduced some warnings about implicit type
> > conversions or whatever, the priority is not so high.
>
> That's where I'm saying we disagree on how critical things are.
> Implicit type conversions breaks the compile, because WARNINGS
> ARE ERRORS. Now, you are a submitter; I'm not. However, in any
> code base I've ever worked, the quality has gone up when technical
> leadership has decided that warnings should break the compile, and
> the switch has been turned on. I am suggesting (although I'm
> failing to be humble about it) that you do the same thing in ODE.
>
> > Actually, ODE compiles with nearly a hundred warnings (MS, dReal=double)
> > and has done so for at least the last year (the time which I've been
> > using it).  I will try to fix all of these and check in the changes.
>
> Interesting. I use dReal == float, and it works fine for me --
> although I think I turned off the "this will break on win64"
> warnings. To my dismay, I might add, but there are some cases
> that requires bigger changes than I'm willing to commit to to
> fix all of that up.
>