cps & contification

[Stephen Weeks]

I found one silly mistake in my definition of safety that could cause
confusion.  Here's the fix.

> An analysis A is safe for program p = (fm, T, N) if 
> 	*   A(fm) = fm
> 	**  For all nontail calls (f, g, j) in N, either
> 		A(f) = Uncalled
> 		or A(f) in {Uncalled, Unknown, j}

should read	or A(g) in {Uncalled, Unknown, j}

> 	*** For all tail calls (f, g) in T, either 
> 		A(f) = Uncalled
> 		or A(g) in {Unknown, f, A(f)}