X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=src%2Fcode%2Fprimordial-extensions.lisp;h=38aa0635b0103963519fed7c83100493d035bdaf;hb=86210c4e406c1b2ff10cc3bac0e71435867db48b;hp=cb45dc3ac169da18a59425a1ae52b733f64b9919;hpb=993d5b779638756473181dda8d928d33038d4cc3;p=sbcl.git diff --git a/src/code/primordial-extensions.lisp b/src/code/primordial-extensions.lisp index cb45dc3..38aa063 100644 --- a/src/code/primordial-extensions.lisp +++ b/src/code/primordial-extensions.lisp @@ -12,23 +12,76 @@ (in-package "SB!INT") +;;;; target constants which need to appear as early as possible + +;;; an internal tag for marking empty slots, which needs to be defined +;;; as early as possible because it appears in macroexpansions for +;;; iteration over hash tables +;;; +;;; CMU CL 18b used :EMPTY for this purpose, which was somewhat nasty +;;; since it's easily accessible to the user, so that e.g. +;;; (DEFVAR *HT* (MAKE-HASH-TABLE)) +;;; (SETF (GETHASH :EMPTY *HT*) :EMPTY) +;;; (MAPHASH (LAMBDA (K V) (FORMAT T "~&~S ~S~%" K V))) +;;; gives no output -- oops! +;;; +;;; FIXME: It'd probably be good to use the unbound marker for this. +;;; However, there might be some gotchas involving assumptions by +;;; e.g. AREF that they're not going to return the unbound marker, +;;; and there's also the noted-below problem that the C-level code +;;; contains implicit assumptions about this marker. +;;; +;;; KLUDGE: Note that as of version 0.pre7 there's a dependence in the +;;; gencgc.c code on this value being a symbol. (This is only one of +;;; several nasty dependencies between that code and this, alas.) +;;; -- WHN 2001-08-17 +;;; +;;; FIXME: We end up doing two DEFCONSTANT forms because (1) LispWorks +;;; needs EVAL-WHEN wrapped around DEFCONSTANT, and (2) SBCL's +;;; DEFCONSTANT expansion doesn't seem to behave properly inside +;;; EVAL-WHEN, so that without this, the +EMPTY-HT-SLOT+ references in +;;; e.g. DOHASH macroexpansions don't end up being replaced by +;;; constant values, so that the system dies at cold init because +;;; '+EMPTY-HT-SLOT+ isn't bound yet. It's hard to fix this properly +;;; until SBCL's EVAL-WHEN is fixed, which is waiting for the IR1 +;;; interpreter to go away, which is waiting for sbcl-0.7.x.. +(eval-when (:compile-toplevel :load-toplevel :execute) + (defconstant +empty-ht-slot+ '%empty-ht-slot%)) +;;; We shouldn't need this mess now that EVAL-WHEN works. +#+nil (defconstant +empty-ht-slot+ '#.+empty-ht-slot+) ; egads.. See FIXME above. +;;; KLUDGE: Using a private symbol still leaves us vulnerable to users +;;; getting nonconforming behavior by messing around with +;;; DO-ALL-SYMBOLS. That seems like a fairly obscure problem, so for +;;; now we just don't worry about it. If for some reason it becomes +;;; worrisome and the magic value needs replacement: +;;; * The replacement value needs to be LOADable with EQL preserved, +;;; so that the macroexpansion for WITH-HASH-TABLE-ITERATOR will +;;; work when compiled into a file and loaded back into SBCL. +;;; (Thus, just uninterning %EMPTY-HT-SLOT% doesn't work.) +;;; * The replacement value needs to be acceptable to the +;;; low-level gencgc.lisp hash table scavenging code. +;;; * The change will break binary compatibility, since comparisons +;;; against the value used at the time of compilation are wired +;;; into FASL files. +;;; -- WHN 20000622 + ;;;; DO-related stuff which needs to be visible on the cross-compilation host (eval-when (:compile-toplevel :load-toplevel :execute) - (defun do-do-body (varlist endlist decls-and-code bind step name block) + (defun frob-do-body (varlist endlist decls-and-code bind step name block) (let* ((r-inits nil) ; accumulator for reversed list (r-steps nil) ; accumulator for reversed list (label-1 (gensym)) (label-2 (gensym))) ;; Check for illegal old-style DO. (when (or (not (listp varlist)) (atom endlist)) - (error "Ill-formed ~S -- possibly illegal old style DO?" name)) + (error "ill-formed ~S -- possibly illegal old style DO?" name)) ;; Parse VARLIST to get R-INITS and R-STEPS. (dolist (v varlist) - (flet (;; (We avoid using CL:PUSH here so that CL:PUSH can be defined - ;; in terms of CL:SETF, and CL:SETF can be defined in terms of - ;; CL:DO, and CL:DO can be defined in terms of the current - ;; function.) + (flet (;; (We avoid using CL:PUSH here so that CL:PUSH can be + ;; defined in terms of CL:SETF, and CL:SETF can be + ;; defined in terms of CL:DO, and CL:DO can be defined + ;; in terms of the current function.) (push-on-r-inits (x) (setq r-inits (cons x r-inits))) ;; common error-handling @@ -41,10 +94,10 @@ name (first v))) (let ((lv (length v))) - ;; (We avoid using CL:CASE here so that CL:CASE can be - ;; defined in terms of CL:SETF, and CL:SETF can be defined - ;; in terms of CL:DO, and CL:DO can be defined in terms of - ;; the current function.) + ;; (We avoid using CL:CASE here so that CL:CASE can + ;; be defined in terms of CL:SETF, and CL:SETF can + ;; be defined in terms of CL:DO, and CL:DO can be + ;; defined in terms of the current function.) (cond ((= lv 1) (push-on-r-inits (first v))) ((= lv 2) @@ -68,8 +121,6 @@ (unless ,(first endlist) (go ,label-1)) (return-from ,block (progn ,@(rest endlist)))))))))) -;;; DO-ANONYMOUS ({(Var [Init] [Step])}*) (Test Exit-Form*) Declaration* Form* -;;; ;;; This is like DO, except it has no implicit NIL block. Each VAR is ;;; initialized in parallel to the value of the specified INIT form. ;;; On subsequent iterations, the VARS are assigned the value of the @@ -78,10 +129,22 @@ ;;; EXIT-FORMS are evaluated as a PROGN, with the result being the ;;; value of the DO. (defmacro do-anonymous (varlist endlist &rest body) - (do-do-body varlist endlist body 'let 'psetq 'do-anonymous (gensym))) + (frob-do-body varlist endlist body 'let 'psetq 'do-anonymous (gensym))) ;;;; miscellany +;;; Lots of code wants to get to the KEYWORD package or the +;;; COMMON-LISP package without a lot of fuss, so we cache them in +;;; variables. TO DO: How much does this actually buy us? It sounds +;;; sensible, but I don't know for sure that it saves space or time.. +;;; -- WHN 19990521 +;;; +;;; (The initialization forms here only matter on the cross-compilation +;;; host; In the target SBCL, these variables are set in cold init.) +(declaim (type package *cl-package* *keyword-package*)) +(defvar *cl-package* (find-package "COMMON-LISP")) +(defvar *keyword-package* (find-package "KEYWORD")) + ;;; Concatenate together the names of some strings and symbols, ;;; producing a symbol in the current package. (eval-when (:compile-toplevel :load-toplevel :execute) @@ -120,12 +183,34 @@ ;; Then complain. (error 'simple-type-error :datum maybe-package - :expected-type 'package + :expected-type '(and package (satisfies package-name)) :format-control - "~@<~S can't be a ~S: ~2I~_~S has been reset to ~S.~:>" - :format-arguments (list '*package* (type-of maybe-package) + "~@<~S can't be a ~A: ~2I~_~S has been reset to ~S.~:>" + :format-arguments (list '*package* + (if (packagep maybe-package) + "deleted package" + (type-of maybe-package)) '*package* really-package))))))) +;;; Access *DEFAULT-PATHNAME-DEFAULTS*, issuing a warning if its value +;;; is silly. (Unlike the vaguely-analogous SANE-PACKAGE, we don't +;;; actually need to reset the variable when it's silly, since even +;;; crazy values of *DEFAULT-PATHNAME-DEFAULTS* don't leave the system +;;; in a state where it's hard to recover interactively.) +(defun sane-default-pathname-defaults () + (let* ((dfd *default-pathname-defaults*) + (dfd-dir (pathname-directory dfd))) + ;; It's generally not good to use a relative pathname for + ;; *DEFAULT-PATHNAME-DEFAULTS*, since relative pathnames + ;; are defined by merging into a default pathname (which is, + ;; by default, *DEFAULT-PATHNAME-DEFAULTS*). + (when (and (consp dfd-dir) + (eql (first dfd-dir) :relative)) + (warn + "~@<~S is a relative pathname. (But we'll try using it anyway.)~@:>" + '*default-pathname-defaults*)) + dfd)) + ;;; Give names to elements of a numeric sequence. (defmacro defenum ((&key (prefix "") (suffix "") (start 0) (step 1)) &rest identifiers) @@ -165,41 +250,25 @@ ;;; structure for each object file which contains code referring to ;;; the value, plus perhaps one more copy bound to the SYMBOL-VALUE of ;;; the constant. If you don't want that to happen, you should -;;; probably use DEFPARAMETER instead. +;;; probably use DEFPARAMETER instead; or if you truly desperately +;;; need to avoid runtime indirection through a symbol, you might be +;;; able to do something with LOAD-TIME-VALUE or MAKE-LOAD-FORM. (defmacro defconstant-eqx (symbol expr eqx &optional doc) - (let ((expr-tmp (gensym "EXPR-TMP-"))) - `(progn - ;; When we're building the cross-compiler, and in most - ;; situations even when we're running the cross-compiler, - ;; all we need is a nice portable definition in terms of the - ;; ANSI Common Lisp operations. - (eval-when (:compile-toplevel :load-toplevel :execute) - (let ((,expr-tmp ,expr)) - (cond ((boundp ',symbol) - (unless (and (constantp ',symbol) - (funcall ,eqx - (symbol-value ',symbol) - ,expr-tmp)) - (error "already bound differently: ~S"))) - (t - (defconstant ,symbol ,expr-tmp ,@(when doc `(,doc))))))) - ;; The #+SB-XC :COMPILE-TOPLEVEL situation is special, since we - ;; want to define the symbol not just in the cross-compilation - ;; host Lisp (which was handled above) but also in the - ;; cross-compiler (which we will handle now). - ;; - ;; KLUDGE: It would probably be possible to do this fairly - ;; cleanly, in a way parallel to the code above, if we had - ;; SB!XC:FOO versions of all the primitives CL:FOO used above - ;; (e.g. SB!XC:BOUNDP, SB!XC:SYMBOL-VALUE, and - ;; SB!XC:DEFCONSTANT), and took care to call them. But right - ;; now we just hack around in the guts of the cross-compiler - ;; instead. -- WHN 2000-11-03 - #+sb-xc - (eval-when (:compile-toplevel) - (let ((,expr-tmp ,symbol)) - (unless (and (eql (info :variable :kind ',symbol) :constant) - (funcall ,eqx - (info :variable :constant-value ',symbol) - ,expr-tmp)) - (sb!c::%defconstant ',symbol ,expr-tmp ,doc))))))) + `(defconstant ,symbol + (%defconstant-eqx-value ',symbol ,expr ,eqx) + ,@(when doc (list doc)))) +(defun %defconstant-eqx-value (symbol expr eqx) + (flet ((bummer (explanation) + (error "~@" + symbol + expr + explanation + (symbol-value symbol)))) + (cond ((not (boundp symbol)) + expr) + ((not (constantp symbol)) + (bummer "already bound as a non-constant")) + ((not (funcall eqx (symbol-value symbol) expr)) + (bummer "already bound as a different constant value")) + (t + (symbol-value symbol)))))