The draft FFI says:

    double SCHEME_EXTRACT_DOUBLE(scheme_value)
    char * SCHEME_EXTRACT_STRING(scheme_value)

Neither says "(may GC)".

I don't like that.

If I'm using some exotic number representation (constructive reals,
perhaps), then EXTRACT_DOUBLE may very well involve some pretty hairy,
hence possibly GC-causing, computation.

If I'm using some exotic string representations (I'm working on a
functional-splay-tree string type for Pika) -- same deal:
extract-string may take some (possibly GC-causing) work.

Even something innocent like:

     int SCHEME_CHAR_P(scheme_value)

can cause GC if my implementation let's me attach to a hook in its
implementation.

In short, even if you are going to be stubborn about concurrent
threads or async execution an insist that GC can only happen "at
certain times" -- I'm not convinced that there is a single entry point
anywhere in the FFI which does not qualify for "(may GC)".

-t