0.pre7.61:
OK, that's enough renaming for a while. I'm tired of rebuilding
taking so long, and from the long compile times of the
DEFSTRUCT-heavy files like node.lisp, the use of
DEFUN instead of closures to define structure
accessors is likely to be contributing, and
that's something I wanted to fix anyway. So in
preparation for removing DEFUNs from DEFSTRUCT
macroexpansion..
..hacked the definition of INFO :FUNCTION :INLINE-EXPANSION
so that it will accept FUNCTION values as well as
lambda expressions, with the nonobvious but convenient
interpretation that the function is to be called
to get a lambda expression.
..wrote FUN-NAME-INLINE-EXPANSION to support this
..renamed other FUNCTION-NAME stuff to have parallel names
..renamed INFO :FUNCTION :INLINE-EXPANSION to
INFO :FUNCTION :INLINE-EXPANSION-DESIGNATOR
..renamed CLASS-STRUCTURE-P to DD-CLASS-P, since I keep
forgetting exactly what CLASS-STRUCTURE-P means
(and vice versa, forgetting the name for this property)
tweaked representation of INFO :VARIABLE :CONSTANT-VALUE so
that it returns only a single value, so that we no
longer need the complexity of VALUES-returning INFO
entries, so that the type declaration of the return
value doesn't wander into the twilight zone of
whether T is a (VALUES T T) and similar questions
that ANSI seems not to've considered
restructured compiler-macro implementation of INFO to
avoid the (VALUES T T) ambiguity
rewrote FIND-FREE-VARIABLE to use bare
(EQL (INFO :VARIABLE :KIND ..) :CONSTANT) instead
of messing with the second value return from
(INFO :VARIABLE :CONSTANT-VALUE ..); and checked that
there are no other uses of the second value
split #'SYMBOL-SELF-EVALUATING-P out of #'ABOUT-TO-MODIFY, and
used it in INFO instead of the funky special casing
of T and NIL in :DEFAULT of INFO :VARIABLE :KIND and
elsewhere
copied Christophe Rhodes' *BACKEND-FEATURES* documentation
from the CLiki SBCL internals site and pasted them
into the source code
50 files changed: