Flat syntactic closures

[mnieper]

Consider the following expression:

(make-syntactic-closure env1 free1 (make-syntactic-closure env2 free2 expr))

Here, env1 and env2 are (syntactic) environments, free1 and free2 are lists of identifiers and expr is an expression. In the syntactic closure (make-syntactic-closure env2 free2 expr) all identifiers in expr except those in free2 shall take their meaning from env2. The identifiers in free2 remain free in the syntactic closure. Thus, in the syntactic closure (make-syntactic-closure env1 free1 (make-syntactic-closure env2 free2 expr), the identifiers in expr take their meaning from env2 if they are not in free2, take their meaning from env1 if they are not in free1 but in free2 and are free in the syntactic closure if they are both in free1 and free2.

Chibi has the feature flag SEXP_USE_FLAT_SYNTACTIC_CLOSURES, which is disabled by default. When enabled, the expression

(make-syntactic-closure env1 free1 (make-syntactic-closure env2 free2 expr))

is evaluated as

(make-syntactic-closure env2 free2 expr)

(see https://github.com/ashinn/chibi-scheme/blob/0efa071672184936741640bb9d06012c7bdbb596/eval.c#L396). Clearly, this has a different meaning than the original expression if free2 is not a sublist of free1 and would violate the specification of make-syntactic-closure. Thus, the code line cited should include a test that free2 is a sublist of free1 before flattening the syntactic closure.

[ashinn]

I'm not sure if the flat syntactic closures are expected to work at all. Did you want to use them?

[mnieper]

No, I just found that code by chance and noticed the error. Don't know whether there are more issues with the flat syntactic closures.

Why did you include that feature?