X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=src%2Fcode%2Ffd-stream.lisp;h=6dff66f750feee883e533ac4dfd74097542592bd;hb=b56c1a4dc22aa0ac827343667584aa6090b15f02;hp=98df2d4db68c8695c4d4351ec3ac94f256278b2a;hpb=a02dc247784acf855dd210660e2c6dbbfafde362;p=sbcl.git diff --git a/src/code/fd-stream.lisp b/src/code/fd-stream.lisp index 98df2d4..6dff66f 100644 --- a/src/code/fd-stream.lisp +++ b/src/code/fd-stream.lisp @@ -52,26 +52,17 @@ #!+sb-doc "List of available buffers.") -(defvar *available-buffers-spinlock* (sb!thread::make-spinlock - :name "lock for *AVAILABLE-BUFFERS*") +(defvar *available-buffers-lock* (sb!thread:make-mutex + :name "lock for *AVAILABLE-BUFFERS*") #!+sb-doc "Mutex for access to *AVAILABLE-BUFFERS*.") (defmacro with-available-buffers-lock ((&optional) &body body) - ;; CALL-WITH-SYSTEM-SPINLOCK because - ;; - ;; 1. streams are low-level enough to be async signal safe, and in - ;; particular a C-c that brings up the debugger while holding the - ;; mutex would lose badly - ;; - ;; 2. this can potentially be a fairly busy (but also probably - ;; uncontended) lock, so we don't want to pay the syscall per - ;; release -- hence a spinlock. - ;; - ;; ...again, once we have smarted locks the spinlock here can become - ;; a mutex. - `(sb!thread::call-with-system-spinlock (lambda () ,@body) - *available-buffers-spinlock*)) + ;; CALL-WITH-SYSTEM-MUTEX because streams are low-level enough to be + ;; async signal safe, and in particular a C-c that brings up the + ;; debugger while holding the mutex would lose badly. + `(sb!thread::with-system-mutex (*available-buffers-lock*) + ,@body)) (defconstant +bytes-per-buffer+ (* 4 1024) #!+sb-doc @@ -157,6 +148,8 @@ (element-type 'base-char) ;; the Unix file descriptor (fd -1 :type fixnum) + ;; What do we know about the FD? + (fd-type :unknown :type keyword) ;; controls when the output buffer is flushed (buffering :full :type (member :full :line :none)) ;; controls whether the input buffer must be cleared before output @@ -169,10 +162,13 @@ (char-pos nil :type (or unsigned-byte null)) ;; T if input is waiting on FD. :EOF if we hit EOF. (listen nil :type (member nil t :eof)) + ;; T if serve-event is allowed when this stream blocks + (serve-events nil :type boolean) ;; the input buffer - (unread nil) + (instead (make-array 0 :element-type 'character :adjustable t :fill-pointer t) :type (array character (*))) (ibuf nil :type (or buffer null)) + (eof-forced-p nil :type (member t nil)) ;; the output buffer (obuf nil :type (or buffer null)) @@ -184,8 +180,14 @@ (timeout nil :type (or single-float null)) ;; pathname of the file this stream is opened to (returned by PATHNAME) (pathname nil :type (or pathname null)) - (external-format :default) - (output-bytes #'ill-out :type function)) + ;; Not :DEFAULT, because we want to match CHAR-SIZE! + (external-format :latin-1) + ;; fixed width, or function to call with a character + (char-size 1 :type (or fixnum function)) + (output-bytes #'ill-out :type function) + ;; a boolean indicating whether the stream is bivalent. For + ;; internal use only. + (bivalent-p nil :type boolean)) (def!method print-object ((fd-stream fd-stream) stream) (declare (type stream stream)) (print-unreadable-object (fd-stream stream :type t :identity t) @@ -256,32 +258,45 @@ (aver (< head tail)) (%queue-and-replace-output-buffer stream)) (t - ;; Try a non-blocking write, queue whatever is left over. + ;; Try a non-blocking write, if SERVE-EVENT is allowed, queue + ;; whatever is left over. Otherwise wait until we can write. (aver (< head tail)) (synchronize-stream-output stream) - (let ((length (- tail head))) - (multiple-value-bind (count errno) - (sb!unix:unix-write (fd-stream-fd stream) (buffer-sap obuf) - head length) - (cond ((eql count length) - ;; Complete write -- we can use the same buffer. - (reset-buffer obuf)) - (count - ;; Partial write -- update buffer status and queue. - ;; Do not use INCF! Another thread might have moved - ;; head... - (setf (buffer-head obuf) (+ count head)) - (%queue-and-replace-output-buffer stream)) - #!-win32 - ((eql errno sb!unix:ewouldblock) - ;; Blocking, queue. - (%queue-and-replace-output-buffer stream)) - (t - (simple-stream-perror "Couldn't write to ~s" - stream errno))))))))))) + (loop + (let ((length (- tail head))) + (multiple-value-bind (count errno) + (sb!unix:unix-write (fd-stream-fd stream) (buffer-sap obuf) + head length) + (flet ((queue-or-wait () + (if (fd-stream-serve-events stream) + (return (%queue-and-replace-output-buffer stream)) + (or (wait-until-fd-usable (fd-stream-fd stream) :output + (fd-stream-timeout stream) + nil) + (signal-timeout 'io-timeout + :stream stream + :direction :output + :seconds (fd-stream-timeout stream)))))) + (cond ((eql count length) + ;; Complete write -- we can use the same buffer. + (return (reset-buffer obuf))) + (count + ;; Partial write -- update buffer status and + ;; queue or wait. + (incf head count) + (setf (buffer-head obuf) head) + (queue-or-wait)) + #!-win32 + ((eql errno sb!unix:ewouldblock) + ;; Blocking, queue or wair. + (queue-or-wait)) + (t + (simple-stream-perror "Couldn't write to ~s" + stream errno))))))))))))) ;;; Helper for FLUSH-OUTPUT-BUFFER -- returns the new buffer. (defun %queue-and-replace-output-buffer (stream) + (aver (fd-stream-serve-events stream)) (let ((queue (fd-stream-output-queue stream)) (later (list (or (fd-stream-obuf stream) (bug "Missing obuf.")))) (new (get-buffer))) @@ -306,6 +321,7 @@ ;;; This is called by the FD-HANDLER for the stream when output is ;;; possible. (defun write-output-from-queue (stream) + (aver (fd-stream-serve-events stream)) (synchronize-stream-output stream) (let (not-first-p) (tagbody @@ -384,14 +400,10 @@ ;;; this is not something we want to export. Nikodemus thinks the ;;; right thing is to support a low-level non-stream like IO layer, ;;; akin to java.nio. -(defun output-raw-bytes (stream thing &optional start end) +(declaim (inline output-raw-bytes)) +(define-deprecated-function :late "1.0.8.16" output-raw-bytes write-sequence + (stream thing &optional start end) (write-or-buffer-output stream thing (or start 0) (or end (length thing)))) - -(define-compiler-macro output-raw-bytes (stream thing &optional start end) - (deprecation-warning 'output-raw-bytes) - (let ((x (gensym "THING"))) - `(let ((,x ,thing)) - (write-or-buffer-output ,stream ,x (or ,start 0) (or ,end (length ,x)))))) ;;;; output routines and related noise @@ -416,12 +428,14 @@ (defun stream-decoding-error (stream octets) (error 'stream-decoding-error + :external-format (stream-external-format stream) :stream stream ;; FIXME: dunno how to get at OCTETS currently, or even if ;; that's the right thing to report. :octets octets)) (defun stream-encoding-error (stream code) (error 'stream-encoding-error + :external-format (stream-external-format stream) :stream stream :code code)) @@ -448,14 +462,31 @@ (attempt-resync () :report (lambda (stream) (format stream - "~@")) (fd-stream-resync stream) nil) (force-end-of-file () :report (lambda (stream) (format stream "~@")) - t))) + (setf (fd-stream-eof-forced-p stream) t)) + (input-replacement (string) + :report (lambda (stream) + (format stream "~@")) + :interactive (lambda () + (format *query-io* "~@") + (finish-output *query-io*) + (list (read *query-io*))) + (let ((string (reverse (string string))) + (instead (fd-stream-instead stream))) + (dotimes (i (length string)) + (vector-push-extend (char string i) instead)) + (fd-stream-resync stream) + (when (> (length string) 0) + (setf (fd-stream-listen stream) t))) + nil))) (defun stream-encoding-error-and-handle (stream code) (restart-case @@ -463,6 +494,16 @@ (output-nothing () :report (lambda (stream) (format stream "~@")) + (throw 'output-nothing nil)) + (output-replacement (string) + :report (lambda (stream) + (format stream "~@")) + :interactive (lambda () + (format *query-io* "~@") + (finish-output *query-io*) + (list (read *query-io*))) + (let ((string (string string))) + (fd-sout stream (string string) 0 (length string))) (throw 'output-nothing nil)))) (defun external-format-encoding-error (stream code) @@ -470,11 +511,6 @@ (stream-encoding-error-and-handle stream code) (c-string-encoding-error stream code))) -(defun external-format-decoding-error (stream octet-count) - (if (streamp stream) - (stream-decoding-error stream octet-count) - (c-string-decoding-error stream octet-count))) - (defun synchronize-stream-output (stream) ;; If we're reading and writing on the same file, flush buffered ;; input and rewind file position accordingly. @@ -695,7 +731,7 @@ (position #\newline thing :from-end t :start start :end end)))) (if (and (typep thing 'base-string) - (eq (fd-stream-external-format stream) :latin-1)) + (eq (fd-stream-external-format-keyword stream) :latin-1)) (ecase (fd-stream-buffering stream) (:full (buffer-output stream thing start end)) @@ -716,38 +752,131 @@ (setf (fd-stream-char-pos stream) (- end last-newline 1)) (incf (fd-stream-char-pos stream) (- end start)))))) -(defvar *external-formats* () +(defstruct (external-format + (:constructor %make-external-format) + (:conc-name ef-) + (:predicate external-format-p) + (:copier %copy-external-format)) + ;; All the names that can refer to this external format. The first + ;; one is the canonical name. + (names (missing-arg) :type list :read-only t) + (default-replacement-character (missing-arg) :type character) + (read-n-chars-fun (missing-arg) :type function) + (read-char-fun (missing-arg) :type function) + (write-n-bytes-fun (missing-arg) :type function) + (write-char-none-buffered-fun (missing-arg) :type function) + (write-char-line-buffered-fun (missing-arg) :type function) + (write-char-full-buffered-fun (missing-arg) :type function) + ;; Can be nil for fixed-width formats. + (resync-fun nil :type (or function null)) + (bytes-for-char-fun (missing-arg) :type function) + (read-c-string-fun (missing-arg) :type function) + (write-c-string-fun (missing-arg) :type function) + ;; We indirect through symbols in these functions so that a + ;; developer working on the octets code can easily redefine things + ;; and use the new function definition without redefining the + ;; external format as well. The slots above don't do any + ;; indirection because a developer working with those slots would be + ;; redefining the external format anyway. + (octets-to-string-fun (missing-arg) :type function) + (string-to-octets-fun (missing-arg) :type function)) + +(defun ef-char-size (ef-entry) + (if (variable-width-external-format-p ef-entry) + (bytes-for-char-fun ef-entry) + (funcall (bytes-for-char-fun ef-entry) #\x))) + +(defun wrap-external-format-functions (external-format fun) + (let ((result (%copy-external-format external-format))) + (macrolet ((frob (accessor) + `(setf (,accessor result) (funcall fun (,accessor result))))) + (frob ef-read-n-chars-fun) + (frob ef-read-char-fun) + (frob ef-write-n-bytes-fun) + (frob ef-write-char-none-buffered-fun) + (frob ef-write-char-line-buffered-fun) + (frob ef-write-char-full-buffered-fun) + (frob ef-resync-fun) + (frob ef-bytes-for-char-fun) + (frob ef-read-c-string-fun) + (frob ef-write-c-string-fun) + (frob ef-octets-to-string-fun) + (frob ef-string-to-octets-fun)) + result)) + +(defvar *external-formats* (make-hash-table) #!+sb-doc - "List of all available external formats. Each element is a list of the - element-type, string input function name, character input function name, - and string output function name.") + "Hashtable of all available external formats. The table maps from + external-format names to EXTERNAL-FORMAT structures.") (defun get-external-format (external-format) - (dolist (entry *external-formats*) - (when (member external-format (first entry)) - (return entry)))) - -(defun get-external-format-function (external-format index) - (let ((entry (get-external-format external-format))) - (when entry (nth index entry)))) + (flet ((keyword-external-format (keyword) + (declare (type keyword keyword)) + (gethash keyword *external-formats*)) + (replacement-handlerify (entry replacement) + (when entry + (wrap-external-format-functions + entry + (lambda (fun) + (and fun + (lambda (&rest rest) + (declare (dynamic-extent rest)) + (handler-bind + ((stream-decoding-error + (lambda (c) + (declare (ignore c)) + (invoke-restart 'input-replacement replacement))) + (stream-encoding-error + (lambda (c) + (declare (ignore c)) + (invoke-restart 'output-replacement replacement))) + (octets-encoding-error + (lambda (c) (use-value replacement c))) + (octet-decoding-error + (lambda (c) (use-value replacement c)))) + (apply fun rest))))))))) + (typecase external-format + (keyword (keyword-external-format external-format)) + ((cons keyword) + (let ((entry (keyword-external-format (car external-format))) + (replacement (getf (cdr external-format) :replacement))) + (if replacement + (replacement-handlerify entry replacement) + entry)))))) + +(defun get-external-format-or-lose (external-format) + (or (get-external-format external-format) + (error "Undefined external-format: ~S" external-format))) + +(defun external-format-keyword (external-format) + (typecase external-format + (keyword external-format) + ((cons keyword) (car external-format)))) + +(defun fd-stream-external-format-keyword (stream) + (external-format-keyword (fd-stream-external-format stream))) + +(defun canonize-external-format (external-format entry) + (typecase external-format + (keyword (first (ef-names entry))) + ((cons keyword) (cons (first (ef-names entry)) (rest external-format))))) ;;; Find an output routine to use given the type and buffering. Return ;;; as multiple values the routine, the real type transfered, and the ;;; number of bytes per element. (defun pick-output-routine (type buffering &optional external-format) (when (subtypep type 'character) - (let ((entry (get-external-format external-format))) - (when entry - (return-from pick-output-routine - (values (symbol-function (nth (ecase buffering - (:none 4) - (:line 5) - (:full 6)) - entry)) - 'character - 1 - (symbol-function (fourth entry)) - (first (first entry))))))) + (let ((entry (get-external-format-or-lose external-format))) + (return-from pick-output-routine + (values (ecase buffering + (:none (ef-write-char-none-buffered-fun entry)) + (:line (ef-write-char-line-buffered-fun entry)) + (:full (ef-write-char-full-buffered-fun entry))) + 'character + 1 + (ef-write-n-bytes-fun entry) + (ef-char-size entry) + (canonize-external-format external-format entry))))) (dolist (entry *output-routines*) (when (and (subtypep type (first entry)) (eq buffering (second entry)) @@ -819,37 +948,25 @@ ;;; correct on win32. However, none of the places that use it require ;;; further assurance than "may" versus "will definitely not". (defun sysread-may-block-p (stream) - #+win32 + #!+win32 ;; This answers T at EOF on win32, I think. (not (sb!win32:fd-listen (fd-stream-fd stream))) - #-win32 - (sb!unix:with-restarted-syscall (count errno) - (sb!alien:with-alien ((read-fds (sb!alien:struct sb!unix:fd-set))) - (sb!unix:fd-zero read-fds) - (sb!unix:fd-set (fd-stream-fd stream) read-fds) - (sb!unix:unix-fast-select (1+ (fd-stream-fd stream)) - (sb!alien:addr read-fds) - nil nil 0 0)) - (case count - ((1) nil) - ((0) t) - (otherwise - (simple-stream-perror "couldn't check whether ~S is readable" - stream - errno))))) + #!-win32 + (not (sb!unix:unix-simple-poll (fd-stream-fd stream) :input 0))) ;;; If the read would block wait (using SERVE-EVENT) till input is available, ;;; then fill the input buffer, and return the number of bytes read. Throws ;;; to EOF-INPUT-CATCHER if the eof was reached. (defun refill-input-buffer (stream) - (let ((fd (fd-stream-fd stream)) - (errno 0) - (count 0)) + (dx-let ((fd (fd-stream-fd stream)) + (errno 0) + (count 0)) (tagbody - ;; Check for blocking input before touching the stream, as if - ;; we happen to wait we are liable to be interrupted, and the - ;; interrupt handler may use the same stream. - (if (sysread-may-block-p stream) + ;; Check for blocking input before touching the stream if we are to + ;; serve events: if the FD is blocking, we don't want to try an uninterruptible + ;; read(). Regular files should never block, so we can elide the check. + (if (and (neq :regular (fd-stream-fd-type stream)) + (sysread-may-block-p stream)) (go :wait-for-input) (go :main)) ;; These (:CLOSED-FLAME and :READ-ERROR) tags are here so what @@ -861,56 +978,71 @@ :wait-for-input ;; This tag is here so we can unwind outside the WITHOUT-INTERRUPTS ;; to wait for input if read tells us EWOULDBLOCK. - (unless (wait-until-fd-usable fd :input (fd-stream-timeout stream)) - (signal-timeout 'io-timeout :stream stream :direction :read + (unless (wait-until-fd-usable fd :input (fd-stream-timeout stream) + (fd-stream-serve-events stream)) + (signal-timeout 'io-timeout + :stream stream + :direction :input :seconds (fd-stream-timeout stream))) :main ;; Since the read should not block, we'll disable the ;; interrupts here, so that we don't accidentally unwind and ;; leave the stream in an inconsistent state. - (without-interrupts - ;; Check the buffer: if it is null, then someone has closed - ;; the stream from underneath us. This is not ment to fix - ;; multithreaded races, but to deal with interrupt handlers - ;; closing the stream. - (let* ((ibuf (or (fd-stream-ibuf stream) (go :closed-flame))) - (sap (buffer-sap ibuf)) - (length (buffer-length ibuf)) - (head (buffer-head ibuf)) - (tail (buffer-tail ibuf))) - (declare (index length head tail)) - (unless (zerop head) - (cond ((eql head tail) - ;; Buffer is empty, but not at yet reset -- make it so. - (setf head 0 - tail 0) - (reset-buffer ibuf)) - (t - ;; Buffer has things in it, but they are not at the head - ;; -- move them there. - (let ((n (- tail head))) - (system-area-ub8-copy sap head sap 0 n) - (setf head 0 - (buffer-head ibuf) head - tail n - (buffer-tail ibuf) tail))))) - (setf (fd-stream-listen stream) nil) - (setf (values count errno) - (sb!unix:unix-read fd (sap+ sap tail) (- length tail))) - (cond ((null count) - #!+win32 - (go :read-error) - #!-win32 - (if (eql errno sb!unix:ewouldblock) - (go :wait-for-input) - (go :read-error))) - ((zerop count) - (setf (fd-stream-listen stream) :eof) - (/show0 "THROWing EOF-INPUT-CATCHER") - (throw 'eof-input-catcher nil)) - (t - ;; Success! (Do not use INCF, for sake of other threads.) - (setf (buffer-tail ibuf) (+ count tail))))))) + + ;; Execute the nlx outside without-interrupts to ensure the + ;; resulting thunk is stack-allocatable. + ((lambda (return-reason) + (ecase return-reason + ((nil)) ; fast path normal cases + ((:wait-for-input) (go :wait-for-input)) + ((:closed-flame) (go :closed-flame)) + ((:read-error) (go :read-error)))) + (without-interrupts + ;; Check the buffer: if it is null, then someone has closed + ;; the stream from underneath us. This is not ment to fix + ;; multithreaded races, but to deal with interrupt handlers + ;; closing the stream. + (block nil + (prog1 nil + (let* ((ibuf (or (fd-stream-ibuf stream) (return :closed-flame))) + (sap (buffer-sap ibuf)) + (length (buffer-length ibuf)) + (head (buffer-head ibuf)) + (tail (buffer-tail ibuf))) + (declare (index length head tail) + (inline sb!unix:unix-read)) + (unless (zerop head) + (cond ((eql head tail) + ;; Buffer is empty, but not at yet reset -- make it so. + (setf head 0 + tail 0) + (reset-buffer ibuf)) + (t + ;; Buffer has things in it, but they are not at the + ;; head -- move them there. + (let ((n (- tail head))) + (system-area-ub8-copy sap head sap 0 n) + (setf head 0 + (buffer-head ibuf) head + tail n + (buffer-tail ibuf) tail))))) + (setf (fd-stream-listen stream) nil) + (setf (values count errno) + (sb!unix:unix-read fd (sap+ sap tail) (- length tail))) + (cond ((null count) + #!+win32 + (return :read-error) + #!-win32 + (if (eql errno sb!unix:ewouldblock) + (return :wait-for-input) + (return :read-error))) + ((zerop count) + (setf (fd-stream-listen stream) :eof) + (/show0 "THROWing EOF-INPUT-CATCHER") + (throw 'eof-input-catcher nil)) + (t + ;; Success! (Do not use INCF, for sake of other threads.) + (setf (buffer-tail ibuf) (+ count tail)))))))))) count)) ;;; Make sure there are at least BYTES number of bytes in the input @@ -937,45 +1069,57 @@ `(let* ((,stream-var ,stream) (ibuf (fd-stream-ibuf ,stream-var)) (size nil)) - (if (fd-stream-unread ,stream-var) - (prog1 - (fd-stream-unread ,stream-var) - (setf (fd-stream-unread ,stream-var) nil) - (setf (fd-stream-listen ,stream-var) nil)) - (let ((,element-var nil) - (decode-break-reason nil)) - (do ((,retry-var t)) - ((not ,retry-var)) - (unless - (catch 'eof-input-catcher - (setf decode-break-reason - (block decode-break-reason - (input-at-least ,stream-var 1) - (let* ((byte (sap-ref-8 (buffer-sap ibuf) - (buffer-head ibuf)))) - (declare (ignorable byte)) - (setq size ,bytes) - (input-at-least ,stream-var size) - (setq ,element-var (locally ,@read-forms)) - (setq ,retry-var nil)) - nil)) - (when decode-break-reason - (stream-decoding-error-and-handle stream - decode-break-reason)) - t) - (let ((octet-count (- (buffer-tail ibuf) - (buffer-head ibuf)))) - (when (or (zerop octet-count) - (and (not ,element-var) - (not decode-break-reason) - (stream-decoding-error-and-handle - stream octet-count))) - (setq ,retry-var nil))))) - (cond (,element-var - (incf (buffer-head ibuf) size) - ,element-var) - (t - (eof-or-lose ,stream-var ,eof-error ,eof-value)))))))) + (block use-instead + (when (fd-stream-eof-forced-p ,stream-var) + (setf (fd-stream-eof-forced-p ,stream-var) nil) + (return-from use-instead + (eof-or-lose ,stream-var ,eof-error ,eof-value))) + (let ((,element-var nil) + (decode-break-reason nil)) + (do ((,retry-var t)) + ((not ,retry-var)) + (if (> (length (fd-stream-instead ,stream-var)) 0) + (let* ((instead (fd-stream-instead ,stream-var)) + (result (vector-pop instead)) + (pointer (fill-pointer instead))) + (when (= pointer 0) + (setf (fd-stream-listen ,stream-var) nil)) + (return-from use-instead result)) + (unless + (catch 'eof-input-catcher + (setf decode-break-reason + (block decode-break-reason + (input-at-least ,stream-var ,(if (consp bytes) (car bytes) `(setq size ,bytes))) + (let* ((byte (sap-ref-8 (buffer-sap ibuf) (buffer-head ibuf)))) + (declare (ignorable byte)) + ,@(when (consp bytes) + `((let ((sap (buffer-sap ibuf)) + (head (buffer-head ibuf))) + (declare (ignorable sap head)) + (setq size ,(cadr bytes)) + (input-at-least ,stream-var size)))) + (setq ,element-var (locally ,@read-forms)) + (setq ,retry-var nil)) + nil)) + (when decode-break-reason + (when (stream-decoding-error-and-handle + stream decode-break-reason) + (setq ,retry-var nil) + (throw 'eof-input-catcher nil))) + t) + (let ((octet-count (- (buffer-tail ibuf) + (buffer-head ibuf)))) + (when (or (zerop octet-count) + (and (not ,element-var) + (not decode-break-reason) + (stream-decoding-error-and-handle + stream octet-count))) + (setq ,retry-var nil)))))) + (cond (,element-var + (incf (buffer-head ibuf) size) + ,element-var) + (t + (eof-or-lose ,stream-var ,eof-error ,eof-value)))))))) ;;; a macro to wrap around all input routines to handle EOF-ERROR noise (defmacro input-wrapper ((stream bytes eof-error eof-value) &body read-forms) @@ -983,11 +1127,8 @@ (element-var (gensym "ELT"))) `(let* ((,stream-var ,stream) (ibuf (fd-stream-ibuf ,stream-var))) - (if (fd-stream-unread ,stream-var) - (prog1 - (fd-stream-unread ,stream-var) - (setf (fd-stream-unread ,stream-var) nil) - (setf (fd-stream-listen ,stream-var) nil)) + (if (> (length (fd-stream-instead ,stream-var)) 0) + (bug "INSTEAD not empty in INPUT-WRAPPER for ~S" ,stream-var) (let ((,element-var (catch 'eof-input-catcher (input-at-least ,stream-var ,bytes) @@ -1073,14 +1214,14 @@ ;;; bytes per element (and for character types string input routine). (defun pick-input-routine (type &optional external-format) (when (subtypep type 'character) - (dolist (entry *external-formats*) - (when (member external-format (first entry)) - (return-from pick-input-routine - (values (symbol-function (third entry)) - 'character - 1 - (symbol-function (second entry)) - (first (first entry))))))) + (let ((entry (get-external-format-or-lose external-format))) + (return-from pick-input-routine + (values (ef-read-char-fun entry) + 'character + 1 + (ef-read-n-chars-fun entry) + (ef-char-size entry) + (canonize-external-format external-format entry))))) (dolist (entry *input-routines*) (when (and (subtypep type (first entry)) (or (not (fourth entry)) @@ -1136,22 +1277,7 @@ &aux (total-copied 0)) (declare (type fd-stream stream)) (declare (type index start requested total-copied)) - (let ((unread (fd-stream-unread stream))) - (when unread - ;; AVERs designed to fail when we have more complicated - ;; character representations. - (aver (typep unread 'base-char)) - (aver (= (fd-stream-element-size stream) 1)) - ;; KLUDGE: this is a slightly-unrolled-and-inlined version of - ;; %BYTE-BLT - (etypecase buffer - (system-area-pointer - (setf (sap-ref-8 buffer start) (char-code unread))) - ((simple-unboxed-array (*)) - (setf (aref buffer start) unread))) - (setf (fd-stream-unread stream) nil) - (setf (fd-stream-listen stream) nil) - (incf total-copied))) + (aver (= (length (fd-stream-instead stream)) 0)) (do () (nil) (let* ((remaining-request (- requested total-copied)) @@ -1183,15 +1309,14 @@ )))) (defun fd-stream-resync (stream) - (dolist (entry *external-formats*) - (when (member (fd-stream-external-format stream) (first entry)) - (return-from fd-stream-resync - (funcall (symbol-function (eighth entry)) stream))))) + (let ((entry (get-external-format (fd-stream-external-format stream)))) + (when entry + (funcall (ef-resync-fun entry) stream)))) (defun get-fd-stream-character-sizer (stream) - (dolist (entry *external-formats*) - (when (member (fd-stream-external-format stream) (first entry)) - (return-from get-fd-stream-character-sizer (ninth entry))))) + (let ((entry (get-external-format (fd-stream-external-format stream)))) + (when entry + (ef-bytes-for-char-fun entry)))) (defun fd-stream-character-size (stream char) (let ((sizer (get-fd-stream-character-sizer stream))) @@ -1204,193 +1329,82 @@ (defun find-external-format (external-format) (when external-format - (find external-format *external-formats* :test #'member :key #'car))) + (get-external-format external-format))) (defun variable-width-external-format-p (ef-entry) - (when (eighth ef-entry) t)) + (and ef-entry (not (null (ef-resync-fun ef-entry))))) (defun bytes-for-char-fun (ef-entry) - (if ef-entry (symbol-function (ninth ef-entry)) (constantly 1))) - -;;; FIXME: OAOOM here vrt. *EXTERNAL-FORMAT-FUNCTIONS* in fd-stream.lisp -(defmacro define-external-format (external-format size output-restart - out-expr in-expr) - (let* ((name (first external-format)) - (out-function (symbolicate "OUTPUT-BYTES/" name)) - (format (format nil "OUTPUT-CHAR-~A-~~A-BUFFERED" (string name))) - (in-function (symbolicate "FD-STREAM-READ-N-CHARACTERS/" name)) - (in-char-function (symbolicate "INPUT-CHAR/" name)) - (size-function (symbolicate "BYTES-FOR-CHAR/" name)) - (read-c-string-function (symbolicate "READ-FROM-C-STRING/" name)) - (output-c-string-function (symbolicate "OUTPUT-TO-C-STRING/" name)) - (n-buffer (gensym "BUFFER"))) + (if ef-entry (ef-bytes-for-char-fun ef-entry) (constantly 1))) + +(defmacro define-unibyte-mapping-external-format + (canonical-name (&rest other-names) &body exceptions) + (let ((->code-name (symbolicate canonical-name '->code-mapper)) + (code->-name (symbolicate 'code-> canonical-name '-mapper)) + (get-bytes-name (symbolicate 'get- canonical-name '-bytes)) + (string->-name (symbolicate 'string-> canonical-name)) + (define-string*-name (symbolicate 'define- canonical-name '->string*)) + (string*-name (symbolicate canonical-name '->string*)) + (define-string-name (symbolicate 'define- canonical-name '->string)) + (string-name (symbolicate canonical-name '->string)) + (->string-aref-name (symbolicate canonical-name '->string-aref))) `(progn - (defun ,size-function (byte) - (declare (ignore byte)) - ,size) - (defun ,out-function (stream string flush-p start end) - (let ((start (or start 0)) - (end (or end (length string)))) - (declare (type index start end)) - (synchronize-stream-output stream) - (unless (<= 0 start end (length string)) - (signal-bounding-indices-bad-error string start end)) - (do () - ((= end start)) - (let ((obuf (fd-stream-obuf stream))) - (setf (buffer-tail obuf) - (string-dispatch (simple-base-string - #!+sb-unicode - (simple-array character (*)) - string) - string - (let ((sap (buffer-sap obuf)) - (len (buffer-length obuf)) - ;; FIXME: rename - (tail (buffer-tail obuf))) - (declare (type index tail) - ;; STRING bounds have already been checked. - (optimize (safety 0))) - (loop - (,@(if output-restart - `(catch 'output-nothing) - `(progn)) - (do* () - ((or (= start end) (< (- len tail) 4))) - (let* ((byte (aref string start)) - (bits (char-code byte))) - ,out-expr - (incf tail ,size) - (incf start))) - ;; Exited from the loop normally - (return tail)) - ;; Exited via CATCH. Skip the current character - ;; and try the inner loop again. - (incf start)))))) - (when (< start end) - (flush-output-buffer stream))) - (when flush-p - (flush-output-buffer stream)))) - (def-output-routines (,format - ,size - ,output-restart - (:none character) - (:line character) - (:full character)) - (if (eql byte #\Newline) - (setf (fd-stream-char-pos stream) 0) - (incf (fd-stream-char-pos stream))) - (let* ((obuf (fd-stream-obuf stream)) - (bits (char-code byte)) - (sap (buffer-sap obuf)) - (tail (buffer-tail obuf))) - ,out-expr)) - (defun ,in-function (stream buffer start requested eof-error-p - &aux (index start) (end (+ start requested))) - (declare (type fd-stream stream) - (type index start requested index end) - (type - (simple-array character (#.+ansi-stream-in-buffer-length+)) - buffer)) - (let ((unread (fd-stream-unread stream))) - (when unread - (setf (aref buffer index) unread) - (setf (fd-stream-unread stream) nil) - (setf (fd-stream-listen stream) nil) - (incf index))) - (do () - (nil) - (let* ((ibuf (fd-stream-ibuf stream)) - (head (buffer-head ibuf)) - (tail (buffer-tail ibuf)) - (sap (buffer-sap ibuf))) - (declare (type index head tail) - (type system-area-pointer sap)) - ;; Copy data from stream buffer into user's buffer. - (dotimes (i (min (truncate (- tail head) ,size) - (- end index))) - (declare (optimize speed)) - (let* ((byte (sap-ref-8 sap head))) - (setf (aref buffer index) ,in-expr) - (incf index) - (incf head ,size))) - (setf (buffer-head ibuf) head) - ;; Maybe we need to refill the stream buffer. - (cond ( ;; If there was enough data in the stream buffer, we're done. - (= index end) - (return (- index start))) - ( ;; If EOF, we're done in another way. - (null (catch 'eof-input-catcher (refill-input-buffer stream))) - (if eof-error-p - (error 'end-of-file :stream stream) - (return (- index start)))) - ;; Otherwise we refilled the stream buffer, so fall - ;; through into another pass of the loop. - )))) - (def-input-routine ,in-char-function (character ,size sap head) - (let ((byte (sap-ref-8 sap head))) - ,in-expr)) - (defun ,read-c-string-function (sap element-type) - (declare (type system-area-pointer sap) - (type (member character base-char) element-type)) - (locally - (declare (optimize (speed 3) (safety 0))) - (let* ((stream ,name) - (length - (loop for head of-type index upfrom 0 by ,size - for count of-type index upto (1- array-dimension-limit) - for byte = (sap-ref-8 sap head) - for char of-type character = ,in-expr - until (zerop (char-code char)) - finally (return count))) - ;; Inline the common cases - (string (make-string length :element-type element-type))) - (declare (ignorable stream) - (type index length) - (type simple-string string)) - (/show0 before-copy-loop) - (loop for head of-type index upfrom 0 by ,size - for index of-type index below length - for byte = (sap-ref-8 sap head) - for char of-type character = ,in-expr - do (setf (aref string index) char)) - string))) ;; last loop rewrite to dotimes? - (defun ,output-c-string-function (string) - (declare (type simple-string string)) - (locally - (declare (optimize (speed 3) (safety 0))) - (let* ((length (length string)) - (,n-buffer (make-array (* (1+ length) ,size) - :element-type '(unsigned-byte 8))) - (tail 0) - (stream ,name)) - (declare (type index length tail)) - (with-pinned-objects (,n-buffer) - (let ((sap (vector-sap ,n-buffer))) - (declare (system-area-pointer sap)) - (dotimes (i length) - (let* ((byte (aref string i)) - (bits (char-code byte))) - (declare (ignorable byte bits)) - ,out-expr) - (incf tail ,size)) - (let* ((bits 0) - (byte (code-char bits))) - (declare (ignorable bits byte)) - ,out-expr))) - ,n-buffer))) - (setf *external-formats* - (cons '(,external-format ,in-function ,in-char-function ,out-function - ,@(mapcar #'(lambda (buffering) - (intern (format nil format (string buffering)))) - '(:none :line :full)) - nil ; no resync-function - ,size-function ,read-c-string-function ,output-c-string-function) - *external-formats*))))) + (define-unibyte-mapper ,->code-name ,code->-name + ,@exceptions) + (declaim (inline ,get-bytes-name)) + (defun ,get-bytes-name (string pos) + (declare (optimize speed (safety 0)) + (type simple-string string) + (type array-range pos)) + (get-latin-bytes #',code->-name ,canonical-name string pos)) + (defun ,string->-name (string sstart send null-padding) + (declare (optimize speed (safety 0)) + (type simple-string string) + (type array-range sstart send)) + (values (string->latin% string sstart send #',get-bytes-name null-padding))) + (defmacro ,define-string*-name (accessor type) + (declare (ignore type)) + (let ((name (make-od-name ',string*-name accessor))) + `(progn + (defun ,name (string sstart send array astart aend) + (,(make-od-name 'latin->string* accessor) + string sstart send array astart aend #',',->code-name))))) + (instantiate-octets-definition ,define-string*-name) + (defmacro ,define-string-name (accessor type) + (declare (ignore type)) + (let ((name (make-od-name ',string-name accessor))) + `(progn + (defun ,name (array astart aend) + (,(make-od-name 'latin->string accessor) + array astart aend #',',->code-name))))) + (instantiate-octets-definition ,define-string-name) + (define-unibyte-external-format ,canonical-name ,other-names + (let ((octet (,code->-name bits))) + (if octet + (setf (sap-ref-8 sap tail) octet) + (external-format-encoding-error stream bits))) + (let ((code (,->code-name byte))) + (if code + (code-char code) + (return-from decode-break-reason 1))) + ,->string-aref-name + ,string->-name)))) + +(defmacro define-unibyte-external-format + (canonical-name (&rest other-names) + out-form in-form octets-to-string-symbol string-to-octets-symbol) + `(define-external-format/variable-width (,canonical-name ,@other-names) + t #\? 1 + ,out-form + 1 + ,in-form + ,octets-to-string-symbol + ,string-to-octets-symbol)) (defmacro define-external-format/variable-width - (external-format output-restart out-size-expr - out-expr in-size-expr in-expr) + (external-format output-restart replacement-character + out-size-expr out-expr in-size-expr in-expr + octets-to-string-sym string-to-octets-sym) (let* ((name (first external-format)) (out-function (symbolicate "OUTPUT-BYTES/" name)) (format (format nil "OUTPUT-CHAR-~A-~~A-BUFFERED" (string name))) @@ -1411,40 +1425,37 @@ (declare (type index start end)) (synchronize-stream-output stream) (unless (<= 0 start end (length string)) - (signal-bounding-indices-bad-error string start end)) + (sequence-bounding-indices-bad-error string start end)) (do () ((= end start)) (let ((obuf (fd-stream-obuf stream))) - (setf (buffer-tail obuf) - (string-dispatch (simple-base-string - #!+sb-unicode - (simple-array character (*)) - string) - string - (let ((len (buffer-length obuf)) - (sap (buffer-sap obuf)) - ;; FIXME: Rename - (tail (buffer-tail obuf))) - (declare (type index tail) - ;; STRING bounds have already been checked. - (optimize (safety 0))) - (loop - (,@(if output-restart - `(catch 'output-nothing) - `(progn)) - (do* () - ((or (= start end) (< (- len tail) 4))) - (let* ((byte (aref string start)) - (bits (char-code byte)) - (size ,out-size-expr)) - ,out-expr - (incf tail size) - (incf start))) - ;; Exited from the loop normally - (return tail)) - ;; Exited via CATCH. Skip the current character - ;; and try the inner loop again. - (incf start)))))) + (string-dispatch (simple-base-string + #!+sb-unicode (simple-array character (*)) + string) + string + (let ((len (buffer-length obuf)) + (sap (buffer-sap obuf)) + ;; FIXME: Rename + (tail (buffer-tail obuf))) + (declare (type index tail) + ;; STRING bounds have already been checked. + (optimize (safety 0))) + (,@(if output-restart + `(catch 'output-nothing) + `(progn)) + (do* () + ((or (= start end) (< (- len tail) 4))) + (let* ((byte (aref string start)) + (bits (char-code byte)) + (size ,out-size-expr)) + ,out-expr + (incf tail size) + (setf (buffer-tail obuf) tail) + (incf start))) + (go flush)) + ;; Exited via CATCH: skip the current character. + (incf start)))) + flush (when (< start end) (flush-output-buffer stream))) (when flush-p @@ -1470,12 +1481,18 @@ (type (simple-array character (#.+ansi-stream-in-buffer-length+)) buffer)) - (let ((unread (fd-stream-unread stream))) - (when unread - (setf (aref buffer start) unread) - (setf (fd-stream-unread stream) nil) - (setf (fd-stream-listen stream) nil) - (incf total-copied))) + (when (fd-stream-eof-forced-p stream) + (setf (fd-stream-eof-forced-p stream) nil) + (return-from ,in-function 0)) + (do ((instead (fd-stream-instead stream))) + ((= (fill-pointer instead) 0) + (setf (fd-stream-listen stream) nil)) + (setf (aref buffer (+ start total-copied)) (vector-pop instead)) + (incf total-copied) + (when (= requested total-copied) + (when (= (fill-pointer instead) 0) + (setf (fd-stream-listen stream) nil)) + (return-from ,in-function total-copied))) (do () (nil) (let* ((ibuf (fd-stream-ibuf stream)) @@ -1489,9 +1506,12 @@ ((or (= tail head) (= requested total-copied))) (setf decode-break-reason (block decode-break-reason + ,@(when (consp in-size-expr) + `((when (> ,(car in-size-expr) (- tail head)) + (return)))) (let ((byte (sap-ref-8 sap head))) (declare (ignorable byte)) - (setq size ,in-size-expr) + (setq size ,(if (consp in-size-expr) (cadr in-size-expr) in-size-expr)) (when (> size (- tail head)) (return)) (setf (aref buffer (+ start total-copied)) ,in-expr) @@ -1513,12 +1533,14 @@ (if eof-error-p (error 'end-of-file :stream stream) (return-from ,in-function total-copied))) - (setf head (buffer-head ibuf)) - (setf tail (buffer-tail ibuf)))) + ;; we might have been given stuff to use instead, so + ;; we have to return (and trust our caller to know + ;; what to do about TOTAL-COPIED being 0). + (return-from ,in-function total-copied))) (setf (buffer-head ibuf) head) ;; Maybe we need to refill the stream buffer. - (cond ( ;; If there were enough data in the stream buffer, we're done. - (= total-copied requested) + (cond ( ;; If was data in the stream buffer, we're done. + (plusp total-copied) (return total-copied)) ( ;; If EOF, we're done in another way. (or (eq decode-break-reason 'eof) @@ -1538,21 +1560,24 @@ (declare (ignorable byte)) ,in-expr)) (defun ,resync-function (stream) - (let ((ibuf (fd-stream-ibuf stream))) - (loop - (input-at-least stream 2) - (incf (buffer-head ibuf)) - (unless (block decode-break-reason - (let* ((sap (buffer-sap ibuf)) - (head (buffer-head ibuf)) - (byte (sap-ref-8 sap head)) - (size ,in-size-expr)) - (declare (ignorable byte)) - (input-at-least stream size) - (setf head (buffer-head ibuf)) - ,in-expr) - nil) - (return))))) + (let ((ibuf (fd-stream-ibuf stream)) + size) + (catch 'eof-input-catcher + (loop + (incf (buffer-head ibuf)) + (input-at-least stream ,(if (consp in-size-expr) (car in-size-expr) `(setq size ,in-size-expr))) + (unless (block decode-break-reason + (let* ((sap (buffer-sap ibuf)) + (head (buffer-head ibuf)) + (byte (sap-ref-8 sap head))) + (declare (ignorable byte)) + ,@(when (consp in-size-expr) + `((setq size ,(cadr in-size-expr)) + (input-at-least stream size))) + (setf head (buffer-head ibuf)) + ,in-expr) + nil) + (return)))))) (defun ,read-c-string-function (sap element-type) (declare (type system-area-pointer sap)) (locally @@ -1564,7 +1589,7 @@ (setf decode-break-reason (block decode-break-reason (setf byte (sap-ref-8 sap head) - size ,in-size-expr + size ,(if (consp in-size-expr) (cadr in-size-expr) in-size-expr) char ,in-expr) (incf head size) nil)) @@ -1583,7 +1608,7 @@ (setf decode-break-reason (block decode-break-reason (setf byte (sap-ref-8 sap head) - size ,in-size-expr + size ,(if (consp in-size-expr) (cadr in-size-expr) in-size-expr) char ,in-expr) (incf head size) nil)) @@ -1632,141 +1657,28 @@ ,out-expr))) ,n-buffer))) - (setf *external-formats* - (cons '(,external-format ,in-function ,in-char-function ,out-function - ,@(mapcar #'(lambda (buffering) - (intern (format nil format (string buffering)))) - '(:none :line :full)) - ,resync-function - ,size-function ,read-c-string-function ,output-c-string-function) - *external-formats*))))) - -;;; Multiple names for the :ISO{,-}8859-* families are needed because on -;;; FreeBSD (and maybe other BSD systems), nl_langinfo("LATIN-1") will -;;; return "ISO8859-1" instead of "ISO-8859-1". -(define-external-format (:latin-1 :latin1 :iso-8859-1 :iso8859-1) - 1 t - (if (>= bits 256) - (external-format-encoding-error stream bits) - (setf (sap-ref-8 sap tail) bits)) - (code-char byte)) - -(define-external-format (:ascii :us-ascii :ansi_x3.4-1968 - :iso-646 :iso-646-us :|646|) - 1 t - (if (>= bits 128) - (external-format-encoding-error stream bits) - (setf (sap-ref-8 sap tail) bits)) - (code-char byte)) - -(let* ((table (let ((s (make-string 256))) - (map-into s #'code-char - '(#x00 #x01 #x02 #x03 #x9c #x09 #x86 #x7f #x97 #x8d #x8e #x0b #x0c #x0d #x0e #x0f - #x10 #x11 #x12 #x13 #x9d #x85 #x08 #x87 #x18 #x19 #x92 #x8f #x1c #x1d #x1e #x1f - #x80 #x81 #x82 #x83 #x84 #x0a #x17 #x1b #x88 #x89 #x8a #x8b #x8c #x05 #x06 #x07 - #x90 #x91 #x16 #x93 #x94 #x95 #x96 #x04 #x98 #x99 #x9a #x9b #x14 #x15 #x9e #x1a - #x20 #xa0 #xe2 #xe4 #xe0 #xe1 #xe3 #xe5 #xe7 #xf1 #xa2 #x2e #x3c #x28 #x2b #x7c - #x26 #xe9 #xea #xeb #xe8 #xed #xee #xef #xec #xdf #x21 #x24 #x2a #x29 #x3b #xac - #x2d #x2f #xc2 #xc4 #xc0 #xc1 #xc3 #xc5 #xc7 #xd1 #xa6 #x2c #x25 #x5f #x3e #x3f - #xf8 #xc9 #xca #xcb #xc8 #xcd #xce #xcf #xcc #x60 #x3a #x23 #x40 #x27 #x3d #x22 - #xd8 #x61 #x62 #x63 #x64 #x65 #x66 #x67 #x68 #x69 #xab #xbb #xf0 #xfd #xfe #xb1 - #xb0 #x6a #x6b #x6c #x6d #x6e #x6f #x70 #x71 #x72 #xaa #xba #xe6 #xb8 #xc6 #xa4 - #xb5 #x7e #x73 #x74 #x75 #x76 #x77 #x78 #x79 #x7a #xa1 #xbf #xd0 #xdd #xde #xae - #x5e #xa3 #xa5 #xb7 #xa9 #xa7 #xb6 #xbc #xbd #xbe #x5b #x5d #xaf #xa8 #xb4 #xd7 - #x7b #x41 #x42 #x43 #x44 #x45 #x46 #x47 #x48 #x49 #xad #xf4 #xf6 #xf2 #xf3 #xf5 - #x7d #x4a #x4b #x4c #x4d #x4e #x4f #x50 #x51 #x52 #xb9 #xfb #xfc #xf9 #xfa #xff - #x5c #xf7 #x53 #x54 #x55 #x56 #x57 #x58 #x59 #x5a #xb2 #xd4 #xd6 #xd2 #xd3 #xd5 - #x30 #x31 #x32 #x33 #x34 #x35 #x36 #x37 #x38 #x39 #xb3 #xdb #xdc #xd9 #xda #x9f)) - s)) - (reverse-table (let ((rt (make-array 256 :element-type '(unsigned-byte 8) :initial-element 0))) - (loop for char across table for i from 0 - do (aver (= 0 (aref rt (char-code char)))) - do (setf (aref rt (char-code char)) i)) - rt))) - (define-external-format (:ebcdic-us :ibm-037 :ibm037) - 1 t - (if (>= bits 256) - (external-format-encoding-error stream bits) - (setf (sap-ref-8 sap tail) (aref reverse-table bits))) - (aref table byte))) - - -#!+sb-unicode -(let ((latin-9-table (let ((table (make-string 256))) - (do ((i 0 (1+ i))) - ((= i 256)) - (setf (aref table i) (code-char i))) - (setf (aref table #xa4) (code-char #x20ac)) - (setf (aref table #xa6) (code-char #x0160)) - (setf (aref table #xa8) (code-char #x0161)) - (setf (aref table #xb4) (code-char #x017d)) - (setf (aref table #xb8) (code-char #x017e)) - (setf (aref table #xbc) (code-char #x0152)) - (setf (aref table #xbd) (code-char #x0153)) - (setf (aref table #xbe) (code-char #x0178)) - table)) - (latin-9-reverse-1 (make-array 16 - :element-type '(unsigned-byte 21) - :initial-contents '(#x0160 #x0161 #x0152 #x0153 0 0 0 0 #x0178 0 0 0 #x20ac #x017d #x017e 0))) - (latin-9-reverse-2 (make-array 16 - :element-type '(unsigned-byte 8) - :initial-contents '(#xa6 #xa8 #xbc #xbd 0 0 0 0 #xbe 0 0 0 #xa4 #xb4 #xb8 0)))) - (define-external-format (:latin-9 :latin9 :iso-8859-15 :iso8859-15) - 1 t - (setf (sap-ref-8 sap tail) - (if (< bits 256) - (if (= bits (char-code (aref latin-9-table bits))) - bits - (external-format-encoding-error stream byte)) - (if (= (aref latin-9-reverse-1 (logand bits 15)) bits) - (aref latin-9-reverse-2 (logand bits 15)) - (external-format-encoding-error stream byte)))) - (aref latin-9-table byte))) - -(define-external-format/variable-width (:utf-8 :utf8) nil - (let ((bits (char-code byte))) - (cond ((< bits #x80) 1) - ((< bits #x800) 2) - ((< bits #x10000) 3) - (t 4))) - (ecase size - (1 (setf (sap-ref-8 sap tail) bits)) - (2 (setf (sap-ref-8 sap tail) (logior #xc0 (ldb (byte 5 6) bits)) - (sap-ref-8 sap (+ 1 tail)) (logior #x80 (ldb (byte 6 0) bits)))) - (3 (setf (sap-ref-8 sap tail) (logior #xe0 (ldb (byte 4 12) bits)) - (sap-ref-8 sap (+ 1 tail)) (logior #x80 (ldb (byte 6 6) bits)) - (sap-ref-8 sap (+ 2 tail)) (logior #x80 (ldb (byte 6 0) bits)))) - (4 (setf (sap-ref-8 sap tail) (logior #xf0 (ldb (byte 3 18) bits)) - (sap-ref-8 sap (+ 1 tail)) (logior #x80 (ldb (byte 6 12) bits)) - (sap-ref-8 sap (+ 2 tail)) (logior #x80 (ldb (byte 6 6) bits)) - (sap-ref-8 sap (+ 3 tail)) (logior #x80 (ldb (byte 6 0) bits))))) - (cond ((< byte #x80) 1) - ((< byte #xc2) (return-from decode-break-reason 1)) - ((< byte #xe0) 2) - ((< byte #xf0) 3) - (t 4)) - (code-char (ecase size - (1 byte) - (2 (let ((byte2 (sap-ref-8 sap (1+ head)))) - (unless (<= #x80 byte2 #xbf) - (return-from decode-break-reason 2)) - (dpb byte (byte 5 6) byte2))) - (3 (let ((byte2 (sap-ref-8 sap (1+ head))) - (byte3 (sap-ref-8 sap (+ 2 head)))) - (unless (and (<= #x80 byte2 #xbf) - (<= #x80 byte3 #xbf)) - (return-from decode-break-reason 3)) - (dpb byte (byte 4 12) (dpb byte2 (byte 6 6) byte3)))) - (4 (let ((byte2 (sap-ref-8 sap (1+ head))) - (byte3 (sap-ref-8 sap (+ 2 head))) - (byte4 (sap-ref-8 sap (+ 3 head)))) - (unless (and (<= #x80 byte2 #xbf) - (<= #x80 byte3 #xbf) - (<= #x80 byte4 #xbf)) - (return-from decode-break-reason 4)) - (dpb byte (byte 3 18) - (dpb byte2 (byte 6 12) - (dpb byte3 (byte 6 6) byte4)))))))) + (let ((entry (%make-external-format + :names ',external-format + :default-replacement-character ,replacement-character + :read-n-chars-fun #',in-function + :read-char-fun #',in-char-function + :write-n-bytes-fun #',out-function + ,@(mapcan #'(lambda (buffering) + (list (intern (format nil "WRITE-CHAR-~A-BUFFERED-FUN" buffering) :keyword) + `#',(intern (format nil format (string buffering))))) + '(:none :line :full)) + :resync-fun #',resync-function + :bytes-for-char-fun #',size-function + :read-c-string-fun #',read-c-string-function + :write-c-string-fun #',output-c-string-function + :octets-to-string-fun (lambda (&rest rest) + (declare (dynamic-extent rest)) + (apply ',octets-to-string-sym rest)) + :string-to-octets-fun (lambda (&rest rest) + (declare (dynamic-extent rest)) + (apply ',string-to-octets-sym rest))))) + (dolist (ef ',external-format) + (setf (gethash ef *external-formats*) entry)))))) ;;;; utility functions (misc routines, etc) @@ -1783,6 +1695,7 @@ (character-stream-p (subtypep target-type 'character)) (bivalent-stream-p (eq element-type :default)) normalized-external-format + char-size (bin-routine #'ill-bin) (bin-type nil) (bin-size nil) @@ -1826,24 +1739,23 @@ (when output-p (setf (fd-stream-char-pos fd-stream) 0)) - (when (and character-stream-p - (eq external-format :default)) + (when (and character-stream-p (eq external-format :default)) (/show0 "/getting default external format") (setf external-format (default-external-format))) (when input-p (when (or (not character-stream-p) bivalent-stream-p) - (multiple-value-setq (bin-routine bin-type bin-size read-n-characters - normalized-external-format) - (pick-input-routine (if bivalent-stream-p '(unsigned-byte 8) - target-type) - external-format)) + (setf (values bin-routine bin-type bin-size read-n-characters + char-size normalized-external-format) + (pick-input-routine (if bivalent-stream-p '(unsigned-byte 8) + target-type) + external-format)) (unless bin-routine (error "could not find any input routine for ~S" target-type))) (when character-stream-p - (multiple-value-setq (cin-routine cin-type cin-size read-n-characters - normalized-external-format) - (pick-input-routine target-type external-format)) + (setf (values cin-routine cin-type cin-size read-n-characters + char-size normalized-external-format) + (pick-input-routine target-type external-format)) (unless cin-routine (error "could not find any input routine for ~S" target-type))) (setf (fd-stream-in fd-stream) cin-routine @@ -1852,8 +1764,8 @@ (setf input-size (or cin-size bin-size)) (setf input-type (or cin-type bin-type)) (when normalized-external-format - (setf (fd-stream-external-format fd-stream) - normalized-external-format)) + (setf (fd-stream-external-format fd-stream) normalized-external-format + (fd-stream-char-size fd-stream) char-size)) (when (= (or cin-size 1) (or bin-size 1) 1) (setf (fd-stream-n-bin fd-stream) ;XXX (if (and character-stream-p (not bivalent-stream-p)) @@ -1880,30 +1792,33 @@ (when output-p (when (or (not character-stream-p) bivalent-stream-p) - (multiple-value-setq (bout-routine bout-type bout-size output-bytes - normalized-external-format) - (pick-output-routine (if bivalent-stream-p - '(unsigned-byte 8) - target-type) - (fd-stream-buffering fd-stream) - external-format)) + (setf (values bout-routine bout-type bout-size output-bytes + char-size normalized-external-format) + (let ((buffering (fd-stream-buffering fd-stream))) + (if bivalent-stream-p + (pick-output-routine '(unsigned-byte 8) + (if (eq :line buffering) + :full + buffering) + external-format) + (pick-output-routine target-type buffering external-format)))) (unless bout-routine (error "could not find any output routine for ~S buffered ~S" (fd-stream-buffering fd-stream) target-type))) (when character-stream-p - (multiple-value-setq (cout-routine cout-type cout-size output-bytes - normalized-external-format) - (pick-output-routine target-type - (fd-stream-buffering fd-stream) - external-format)) + (setf (values cout-routine cout-type cout-size output-bytes + char-size normalized-external-format) + (pick-output-routine target-type + (fd-stream-buffering fd-stream) + external-format)) (unless cout-routine (error "could not find any output routine for ~S buffered ~S" (fd-stream-buffering fd-stream) target-type))) (when normalized-external-format - (setf (fd-stream-external-format fd-stream) - normalized-external-format)) + (setf (fd-stream-external-format fd-stream) normalized-external-format + (fd-stream-char-size fd-stream) char-size)) (when character-stream-p (setf (fd-stream-output-bytes fd-stream) output-bytes)) (setf (fd-stream-out fd-stream) cout-routine @@ -1937,31 +1852,36 @@ input-type output-type)))))) -;;; Handles the resource-release aspects of stream closing. +;;; Handles the resource-release aspects of stream closing, and marks +;;; it as closed. (defun release-fd-stream-resources (fd-stream) (handler-case (without-interrupts + ;; Drop handlers first. + (when (fd-stream-handler fd-stream) + (remove-fd-handler (fd-stream-handler fd-stream)) + (setf (fd-stream-handler fd-stream) nil)) ;; Disable interrupts so that a asynch unwind will not leave ;; us with a dangling finalizer (that would close the same - ;; --possibly reassigned-- FD again). + ;; --possibly reassigned-- FD again), or a stream with a closed + ;; FD that appears open. (sb!unix:unix-close (fd-stream-fd fd-stream)) + (set-closed-flame fd-stream) (when (fboundp 'cancel-finalization) (cancel-finalization fd-stream))) ;; On error unwind from WITHOUT-INTERRUPTS. (serious-condition (e) (error e))) - ;; Release all buffers. If this is undone, or interrupted, ;; we're still safe: buffers have finalizers of their own. (release-fd-stream-buffers fd-stream)) -;;; Flushes the current input buffer and unread chatacter, and returns -;;; the input buffer, and the amount of of flushed input in bytes. +;;; Flushes the current input buffer and any supplied replacements, +;;; and returns the input buffer, and the amount of of flushed input +;;; in bytes. (defun flush-input-buffer (stream) - (let ((unread (if (fd-stream-unread stream) - 1 - 0))) - (setf (fd-stream-unread stream) nil) + (let ((unread (length (fd-stream-instead stream)))) + (setf (fill-pointer (fd-stream-instead stream)) 0) (let ((ibuf (fd-stream-ibuf stream))) (if ibuf (let ((head (buffer-head ibuf)) @@ -2019,70 +1939,80 @@ (do-listen))))))) (do-listen))) (:unread - (setf (fd-stream-unread fd-stream) arg1) - (setf (fd-stream-listen fd-stream) t)) + (decf (buffer-head (fd-stream-ibuf fd-stream)) + (fd-stream-character-size fd-stream arg1))) (:close - (cond (arg1 ; We got us an abort on our hands. - (when (fd-stream-handler fd-stream) - (remove-fd-handler (fd-stream-handler fd-stream)) - (setf (fd-stream-handler fd-stream) nil)) - ;; We can't do anything unless we know what file were - ;; dealing with, and we don't want to do anything - ;; strange unless we were writing to the file. - (when (and (fd-stream-file fd-stream) (fd-stream-obuf fd-stream)) - (if (fd-stream-original fd-stream) - ;; If the original is EQ to file we are appending - ;; and can just close the file without renaming. - (unless (eq (fd-stream-original fd-stream) - (fd-stream-file fd-stream)) - ;; We have a handle on the original, just revert. + ;; Drop input buffers + (setf (ansi-stream-in-index fd-stream) +ansi-stream-in-buffer-length+ + (ansi-stream-cin-buffer fd-stream) nil + (ansi-stream-in-buffer fd-stream) nil) + (cond (arg1 + ;; We got us an abort on our hands. + (let ((outputp (fd-stream-obuf fd-stream)) + (file (fd-stream-file fd-stream)) + (orig (fd-stream-original fd-stream))) + ;; This takes care of the important stuff -- everything + ;; rest is cleaning up the file-system, which we cannot + ;; do on some platforms as long as the file is open. + (release-fd-stream-resources fd-stream) + ;; We can't do anything unless we know what file were + ;; dealing with, and we don't want to do anything + ;; strange unless we were writing to the file. + (when (and outputp file) + (if orig + ;; If the original is EQ to file we are appending to + ;; and can just close the file without renaming. + (unless (eq orig file) + ;; We have a handle on the original, just revert. + (multiple-value-bind (okay err) + (sb!unix:unix-rename orig file) + ;; FIXME: Why is this a SIMPLE-STREAM-ERROR, and the + ;; others are SIMPLE-FILE-ERRORS? Surely they should + ;; all be the same? + (unless okay + (error 'simple-stream-error + :format-control + "~@" + :format-arguments + (list file orig fd-stream (strerror err)) + :stream fd-stream)))) + ;; We can't restore the original, and aren't + ;; appending, so nuke that puppy. + ;; + ;; FIXME: This is currently the fate of superseded + ;; files, and according to the CLOSE spec this is + ;; wrong. However, there seems to be no clean way to + ;; do that that doesn't involve either copying the + ;; data (bad if the :abort resulted from a full + ;; disk), or renaming the old file temporarily + ;; (probably bad because stream opening becomes more + ;; racy). (multiple-value-bind (okay err) - (sb!unix:unix-rename (fd-stream-original fd-stream) - (fd-stream-file fd-stream)) + (sb!unix:unix-unlink file) (unless okay - (simple-stream-perror - "couldn't restore ~S to its original contents" - fd-stream - err)))) - ;; We can't restore the original, and aren't - ;; appending, so nuke that puppy. - ;; - ;; FIXME: This is currently the fate of superseded - ;; files, and according to the CLOSE spec this is - ;; wrong. However, there seems to be no clean way to - ;; do that that doesn't involve either copying the - ;; data (bad if the :abort resulted from a full - ;; disk), or renaming the old file temporarily - ;; (probably bad because stream opening becomes more - ;; racy). - (multiple-value-bind (okay err) - (sb!unix:unix-unlink (fd-stream-file fd-stream)) - (unless okay - (error 'simple-file-error - :pathname (fd-stream-file fd-stream) - :format-control - "~@" - :format-arguments (list (fd-stream-file fd-stream) - (strerror err)))))))) + (error 'simple-file-error + :pathname file + :format-control + "~@" + :format-arguments + (list file fd-stream (strerror err))))))))) (t (finish-fd-stream-output fd-stream) - (when (and (fd-stream-original fd-stream) - (fd-stream-delete-original fd-stream)) - (multiple-value-bind (okay err) - (sb!unix:unix-unlink (fd-stream-original fd-stream)) - (unless okay - (error 'simple-file-error - :pathname (fd-stream-original fd-stream) - :format-control - "~@" - :format-arguments - (list (fd-stream-original fd-stream) - fd-stream - (strerror err)))))))) - (release-fd-stream-resources fd-stream) - ;; Mark as closed. FIXME: Maybe this should be the first thing done? - (sb!impl::set-closed-flame fd-stream)) + (let ((orig (fd-stream-original fd-stream))) + (when (and orig (fd-stream-delete-original fd-stream)) + (multiple-value-bind (okay err) (sb!unix:unix-unlink orig) + (unless okay + (error 'simple-file-error + :pathname orig + :format-control + "~@" + :format-arguments + (list orig fd-stream (strerror err))))))) + ;; In case of no-abort close, don't *really* close the + ;; stream until the last moment -- the cleaning up of the + ;; original can be done first. + (release-fd-stream-resources fd-stream)))) (:clear-input (fd-stream-clear-input fd-stream)) (:force-output @@ -2147,6 +2077,7 @@ (flush-output-buffer stream) (do () ((null (fd-stream-output-queue stream))) + (aver (fd-stream-serve-events stream)) (serve-all-events))) (defun fd-stream-get-file-position (stream) @@ -2176,8 +2107,6 @@ (let ((ibuf (fd-stream-ibuf stream))) (when ibuf (decf posn (- (buffer-tail ibuf) (buffer-head ibuf))))) - (when (fd-stream-unread stream) - (decf posn)) ;; Divide bytes by element size. (truncate posn (fd-stream-element-size stream)))))) @@ -2250,6 +2179,9 @@ ;;; FILE is the name of the file (will be returned by PATHNAME). ;;; ;;; NAME is used to identify the stream when printed. +;;; +;;; If SERVE-EVENTS is true, SERVE-EVENT machinery is used to +;;; handle blocking IO on the stream. (defun make-fd-stream (fd &key (input nil input-p) @@ -2257,6 +2189,7 @@ (element-type 'base-char) (buffering :full) (external-format :default) + serve-events timeout file original @@ -2275,6 +2208,12 @@ ((not (or input output)) (error "File descriptor must be opened either for input or output."))) (let ((stream (%make-fd-stream :fd fd + :fd-type (progn + #!-win32 (sb!unix:fd-type fd) + ;; KLUDGE. + #!+win32 (if serve-events + :unknown + :regular)) :name name :file file :original original @@ -2282,7 +2221,8 @@ :pathname pathname :buffering buffering :dual-channel-p dual-channel-p - :external-format external-format + :bivalent-p (eq element-type :default) + :serve-events serve-events :timeout (if timeout (coerce timeout 'single-float) @@ -2353,19 +2293,23 @@ ;; Calculate useful stuff. (multiple-value-bind (input output mask) - (case direction + (ecase direction (:input (values t nil sb!unix:o_rdonly)) (:output (values nil t sb!unix:o_wronly)) (:io (values t t sb!unix:o_rdwr)) (:probe (values t nil sb!unix:o_rdonly))) (declare (type index mask)) - (let* ((pathname (merge-pathnames filename)) - (namestring - (cond ((unix-namestring pathname input)) - ((and input (eq if-does-not-exist :create)) - (unix-namestring pathname nil)) - ((and (eq direction :io) (not if-does-not-exist-given)) - (unix-namestring pathname nil))))) + (let* (;; PATHNAME is the pathname we associate with the stream. + (pathname (merge-pathnames filename)) + (physical (physicalize-pathname pathname)) + (truename (probe-file physical)) + ;; NAMESTRING is the native namestring we open the file with. + (namestring (cond (truename + (native-namestring truename :as-file t)) + ((or (not input) + (and input (eq if-does-not-exist :create)) + (and (eq direction :io) (not if-does-not-exist-given))) + (native-namestring physical :as-file t))))) ;; Process if-exists argument if we are doing any output. (cond (output (unless if-exists-given @@ -2430,7 +2374,7 @@ (when (and output (= (logand orig-mode #o170000) #o40000)) (error 'simple-file-error - :pathname namestring + :pathname pathname :format-control "can't open ~S for output: is a directory" :format-arguments (list namestring))) @@ -2471,6 +2415,10 @@ (cond ((numberp fd) (case direction ((:input :output :io) + ;; For O_APPEND opened files, lseek returns 0 until first write. + ;; So we jump ahead here. + (when (eq if-exists :append) + (sb!unix:unix-lseek fd 0 sb!unix:l_xtnd)) (make-fd-stream fd :input input :output output @@ -2481,6 +2429,7 @@ :delete-original delete-original :pathname pathname :dual-channel-p nil + :serve-events nil :input-buffer-p t :auto-close t)) (:probe @@ -2537,6 +2486,14 @@ (without-package-locks (makunbound '*available-buffers*)))) +(defun stdstream-external-format (outputp) + (declare (ignorable outputp)) + (let* ((keyword #!+win32 (if outputp (sb!win32::console-output-codepage) (sb!win32::console-input-codepage)) + #!-win32 (default-external-format)) + (ef (get-external-format keyword)) + (replacement (ef-default-replacement-character ef))) + `(,keyword :replacement ,replacement))) + ;;; This is called whenever a saved core is restarted. (defun stream-reinit (&optional init-buffers-p) (when init-buffers-p @@ -2546,22 +2503,25 @@ (with-output-to-string (*error-output*) (setf *stdin* (make-fd-stream 0 :name "standard input" :input t :buffering :line - #!+win32 :external-format #!+win32 (sb!win32::console-input-codepage))) + :element-type :default + :serve-events t + :external-format (stdstream-external-format nil))) (setf *stdout* (make-fd-stream 1 :name "standard output" :output t :buffering :line - #!+win32 :external-format #!+win32 (sb!win32::console-output-codepage))) + :element-type :default + :external-format (stdstream-external-format t))) (setf *stderr* (make-fd-stream 2 :name "standard error" :output t :buffering :line - #!+win32 :external-format #!+win32 (sb!win32::console-output-codepage))) + :element-type :default + :external-format (stdstream-external-format t))) (let* ((ttyname #.(coerce "/dev/tty" 'simple-base-string)) (tty (sb!unix:unix-open ttyname sb!unix:o_rdwr #o666))) (if tty (setf *tty* - (make-fd-stream tty - :name "the terminal" - :input t - :output t - :buffering :line + (make-fd-stream tty :name "the terminal" + :input t :output t :buffering :line + :external-format (stdstream-external-format t) + :serve-events t :auto-close t)) (setf *tty* (make-two-way-stream *stdin* *stdout*)))) (princ (get-output-stream-string *error-output*) *stderr*)) @@ -2583,7 +2543,8 @@ (cond (new-name (setf (fd-stream-pathname stream) new-name) (setf (fd-stream-file stream) - (unix-namestring new-name nil)) + (native-namestring (physicalize-pathname new-name) + :as-file t)) t) (t (fd-stream-pathname stream)))))