-;;; This is like old CMU CL PRE-ENVIRONMENT-ANALYZE-TOPLEVEL, except
-;;; (1) It's been brought into the post-0.7.0 world where the property
-;;; HAS-EXTERNAL-REFERENCES-P is orthogonal to the property of
-;;; being specialized/optimized for locall at top level.
-;;; (2) There's no return value, since we don't care whether we
-;;; find any possible closure variables.
-;;;
-;;; I wish I could find an explanation of why
-;;; PRE-ENVIRONMENT-ANALYZE-TOPLEVEL is important. The old CMU CL
-;;; comments said
-;;; Called on component with top level lambdas before the
-;;; compilation of the associated non-top-level code to detect
-;;; closed over top level variables. We just do COMPUTE-CLOSURE on
-;;; all the lambdas. This will pre-allocate environments for all
-;;; the functions with closed-over top level variables. The
-;;; post-pass will use the existing structure, rather than
-;;; allocating a new one. We return true if we discover any
-;;; possible closure vars.
-;;; But that doesn't seem to explain either why it's important to do
-;;; this for top level lambdas, or why it's important to do it only
-;;; for top level lambdas instead of just doing it indiscriminately
-;;; for all lambdas. I do observe that when it's not done, compiler
-;;; assertions occasionally fail. My tentative hypothesis for why it's
-;;; important to do it is that other environment analysis expects to
-;;; bottom out on the outermost enclosing thing, and (insert
-;;; mysterious reason here) it's important to set up bottomed-out-here
-;;; environments before anything else. I haven't been able to guess
-;;; why it's important to do it selectively instead of
-;;; indiscriminately. -- WHN 2001-11-10
-(defun preallocate-physenvs-for-toplevelish-lambdas (component)
- (dolist (clambda (component-lambdas component))
- (when (lambda-toplevelish-p clambda)
- (add-lambda-vars-and-let-vars-to-closures clambda)))
- (values))
-