On 5/29/07, Donal K. Fellows <donal.k.fellows@[...].uk>  wrote:
> 
>  [...] what you could do instead is to define a new type of
>  variable record that keeps the actual storage space for the variable's
>  contents on the stack (and which [trace] will refuse to touch).


Yeah, that's what I was (tediously, and unknowingly ;-) driving at with my
"hint left in the var hashtable"...

Since all
>  access to variables is through a pretty small API (except in TEBC)
>  making the code work with this is practical.
> 

Hear, hear ! Donal  says it's doable  :-)
I know who you think should do it though ...

-Alex