[Vm-dev] function signatures

[David T. Lewis]

On Wed, Feb 20, 2013 at 10:41:11AM -0800, Eliot Miranda wrote:
>  
> Apologies, David.  I didn't mean to impune you.  I'll try and find
> time to repeat this experiment myself soon.

Sorry, I should have put a smiley on it. I'm afraid that I was rather
snarky sounding now that I re-read it :)

Anyhow, you are offering good advice and I'll try it again myself when
I get some free time (though I've been having fun playing with MVC lately).
I tend to go down a lot of blind alleys working on VMMaker, and that
actually *is* a matter of competence because I'm frequently just figuring
stuff out by fumbling my way through it. Your suggestion is probably
one of the things I tried and gave up on thinking it was a blind alley.

Thanks,
Dave

> 
> On Tue, Feb 19, 2013 at 5:42 PM, David T. Lewis <lewis at mail.msen.com> wrote:
> >
> > On Tue, Feb 19, 2013 at 04:05:00PM -0800, Eliot Miranda wrote:
> >>
> >> On Fri, Feb 15, 2013 at 12:33 PM, David T. Lewis <lewis at mail.msen.com> wrote:
> >> >
> >> > On Fri, Feb 15, 2013 at 09:23:00PM +0100, Noah Evans wrote:
> >> >>
> >> >> On the subject of warnings, is it possible (or are you already
> >> >> starting) to add some sort of escape analysis to VMMaker? The compiler
> >> >> is going ape about functions with no returns and others with returns
> >> >> that are unreachable. I'm happy to do it if people are interested.
> >> >>
> >> >> Noah
> >> >>
> >> >
> >> > I think Eliot has done a good deal of work to correct issues like this
> >> > in the Cog branch, so it's likely that we would see significantly fewer
> >> > of these issues in the code generated by VMMaker in the oscog branch.
> >> > Aside from that, I do not know of anyone doing doing the kind of analysis
> >> > you describe.
> >> >
> >> > I would be very interested in knowing about unreachable returns. The
> >> > functions with no returns are less interesting, because I suspect that
> >> > Eliot has already fixed the underlying code generator issues. In other
> >> > words, as merging of oscog and trunk VMMaker proceeds, I would expect
> >> > that much of this will get cleaned up along the way.
> >>
> >> So the brave thing to do is to use the Cog branch's CCodeGenerator and
> >> TParseNode hierarchy and see what you get.  You should be pleasantly
> >> surprised.
> >
> > I am not lacking in bravery, although my competence may be considerably
> > more in doubt. I'm quite certain that this is one of many things that I
> > have tried.
> >
> > Dave
> >
> 
> 
> 
> -- 
> best,
> Eliot