Yesterday, I was looking at some of our tickets and decided to hack the compiler to support ticket #90 [1]. As with earlier special bindings fixes, I've tested this feature on Maxima which - at least in older versions - is highly sensible to the performance of special variables.

The performance test shows no relevant difference in performance between the clean and the patched version (1140sec patched, 1087sec unpatched - I attribute the difference to statistical variation).

So, the question is: even though the generated code is cleaner and smaller, do we want to make the compiler more complex to support it?

The patch in question is attached for those interested.

Bye,

Erik.

[1] http://trac.common-lisp.net/armedbear/ticket/90