X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=src%2Fcompiler%2Fdump.lisp;h=bb8ff4811b4a04436e3e7600c4b72c35718a7245;hb=cf4cb9554515c59eddbde38d1cf236339c37f55f;hp=991dc1897c5ef2e0e33785b6f65f55a4a4daee48;hpb=670010e3f3dcd62efaf23f61abdc73950edb88c6;p=sbcl.git diff --git a/src/compiler/dump.lisp b/src/compiler/dump.lisp index 991dc18..bb8ff48 100644 --- a/src/compiler/dump.lisp +++ b/src/compiler/dump.lisp @@ -798,9 +798,12 @@ ;; (SIMPLE-ARRAY (UNSIGNED-BYTE 8) *).) The other cases are only ;; needed in the target SBCL, so we let them be handled with ;; unportable bit bashing. - (cond ((>= size 8) ; easy cases + (cond ((>= size 7) ; easy cases (multiple-value-bind (floor rem) (floor size 8) - (aver (zerop rem)) + (aver (or (zerop rem) (= rem 7))) + (when (= rem 7) + (setq size (1+ size)) + (setq floor (1+ floor))) (dovector (i vec) (dump-integer-as-n-bytes (ecase sb!c:*backend-byte-order* @@ -821,6 +824,9 @@ (dump-byte size file)) (dump-raw-bytes vec bytes file))) (etypecase vec + #-sb-xc-host + ((simple-array nil (*)) + (dump-unsigned-vector 0 0)) ;; KLUDGE: What exactly does the (ASH .. -3) stuff do? -- WHN 19990902 (simple-bit-vector (dump-unsigned-vector 1 (ash (+ (the index len) 7) -3))) @@ -837,16 +843,27 @@ #-sb-xc-host ((simple-array (unsigned-byte 4) (*)) (dump-unsigned-vector 4 (ash (+ (the index (ash len 2)) 7) -3))) + #-sb-xc-host + ((simple-array (unsigned-byte 7) (*)) + (dump-unsigned-vector 7 len)) ((simple-array (unsigned-byte 8) (*)) (dump-unsigned-vector 8 len)) + #-sb-xc-host + ((simple-array (unsigned-byte 15) (*)) + (dump-unsigned-vector 15 (* 2 len))) ((simple-array (unsigned-byte 16) (*)) (dump-unsigned-vector 16 (* 2 len))) + #-sb-xc-host + ((simple-array (unsigned-byte 31) (*)) + (dump-unsigned-vector 31 (* 4 len))) ((simple-array (unsigned-byte 32) (*)) (dump-unsigned-vector 32 (* 4 len))) ((simple-array (signed-byte 8) (*)) (dump-signed-vector 8 len)) ((simple-array (signed-byte 16) (*)) (dump-signed-vector 16 (* 2 len))) + ((simple-array (unsigned-byte 29) (*)) + (dump-signed-vector 29 (* 4 len))) ((simple-array (signed-byte 30) (*)) (dump-signed-vector 30 (* 4 len))) ((simple-array (signed-byte 32) (*)) @@ -950,27 +967,18 @@ ;;; - code object references: don't need a name. (defun dump-fixups (fixups fasl-output) (declare (list fixups) (type fasl-output fasl-output)) - (dolist (info fixups) - ;; FIXME: Packing data with LIST in NOTE-FIXUP and unpacking them - ;; with FIRST, SECOND, and THIRD here is hard to follow and - ;; maintain. Perhaps we could define a FIXUP-INFO structure to use - ;; instead, and rename *FIXUPS* to *FIXUP-INFO-LIST*? - (let* ((kind (first info)) - (fixup (second info)) + (dolist (note fixups) + (let* ((kind (fixup-note-kind note)) + (fixup (fixup-note-fixup note)) + (position (fixup-note-position note)) (name (fixup-name fixup)) - (flavor (fixup-flavor fixup)) - (offset (third info))) - ;; FIXME: This OFFSET is not what's called OFFSET in the FIXUP - ;; structure, it's what's called POSN in NOTE-FIXUP. (As far as - ;; I can tell, FIXUP-OFFSET is not actually an offset, it's an - ;; internal label used instead of NAME for :CODE-OBJECT fixups. - ;; Notice that in the :CODE-OBJECT case, NAME is ignored.) + (flavor (fixup-flavor fixup))) (dump-fop 'fop-normal-load fasl-output) (let ((*cold-load-dump* t)) (dump-object kind fasl-output)) (dump-fop 'fop-maybe-cold-load fasl-output) ;; Depending on the flavor, we may have various kinds of - ;; noise before the offset. + ;; noise before the position. (ecase flavor (:assembly-routine (aver (symbolp name)) @@ -990,8 +998,8 @@ (:code-object (aver (null name)) (dump-fop 'fop-code-object-fixup fasl-output))) - ;; No matter what the flavor, we'll always dump the offset. - (dump-unsigned-32 offset fasl-output))) + ;; No matter what the flavor, we'll always dump the position + (dump-unsigned-32 position fasl-output))) (values)) ;;; Dump out the constant pool and code-vector for component, push the @@ -1034,6 +1042,8 @@ ;; hardwired to be empty. And SBCL doesn't have GENGC (and as ;; far as I know no modern CMU CL does either -- WHN ;; 2001-10-05). So might we be able to get rid of trace tables? + ;; + ;; Note that gencgc also does something with the trace table. ;; Dump the constants, noting any :ENTRY constants that have to ;; be patched.