Package: emacs;
Reported by: Ivan Shmakov <ivan <at> siamics.net>
Date: Thu, 4 Dec 2014 21:19:02 UTC
Severity: wishlist
Tags: patch
Fixed in version 25.1
Done: Ivan Shmakov <ivan <at> siamics.net>
Bug is archived. No further changes may be made.
Message #44 received at 19274 <at> debbugs.gnu.org (full text, mbox):
From: Ivan Shmakov <ivan <at> siamics.net> To: 19274 <at> debbugs.gnu.org Subject: Re: bug#19274: tar-mode.el: allow for adding new archive members Date: Sun, 07 Dec 2014 17:47:30 +0000
[Message part 1 (text/plain, inline)]
>>>>> Eli Zaretskii <eliz <at> gnu.org> writes: >>>>> From: Ivan Shmakov Date: Sat, 06 Dec 2014 20:50:36 +0000 >>> fgrep -w Tar doc/emacs/*.texi >> Now, it makes me wonder if there should be an @findex tar-mode along >> with the @cindices? (And @findex archive-mode, too.) > What, just to help us search for it? These functions are never > mentioned in that node. > But I don't mind adding the index entries. I guess my idea of a better Emacs manual includes documenting all the command names alongside their respective default bindings. Looks like it warrants a separate bug report, though. > (In general, the best way of doing this search is to take note of the > node you are reading in Info, and then search for the corresponding > @node line.) In this particular case, I should have tried the command below. $ grep -irE --include=\*.texi -- '\<tar\>' doc/ […] >> The command inserts the new file on the line above the current one, >> so that using it on the topmost line of the Tar buffer makes the new >> file the first one in the archive, and using it on the last line >> (the one after the line describing the last file) makes it the last >> one. > Instead of the complicated description of "the last line, which is > one line after the last file's line", I'd simply say "at the end of > the buffer". This is simpler and much more clear, especially to > non-native speakers who might have problems with long and complex > sentences. Indeed, that’s much better. Does it make sense to also simplify the head of the sentence as follows? (I believe that “lines” are rather just an implementation detail here.) - The command inserts the new file on the line above the current one, + The command inserts the new file before the current one, > Otherwise, it's fine. Thanks. (Don't forget to mark the NEWS entry > with "+++" together with installing this change.) Meanwhile, I’ve also simplified tar-header-serialize a bit (at the cost of adding a couple of lines to tar--put-at) – and hopefully made it more readable at the same time. The (once again) revised patch is MIMEd. -- FSF associate member #7257 http://boycottsystemd.org/ … 3013 B6A0 230E 334A
[Message part 2 (text/diff, inline)]
--- a/etc/NEWS 2014-11-27 11:36:08 +0000
+++ b/etc/NEWS 2014-12-07 17:31:11 +0000
@@ -340,6 +340,10 @@
`tildify-ignored-environments-alist' variables (as well as a few
helper functions) obsolete.
++++
+** tar-mode: new `tar-new-entry' command, allowing for new members to
+be added to the archive.
+
** Obsolete packages
---
--- a/doc/emacs/files.texi 2014-06-08 07:41:27 +0000
+++ b/doc/emacs/files.texi 2014-12-07 17:35:34 +0000
@@ -1689,6 +1689,13 @@ @node File Archives
another window, so you could edit the file and operate on the archive
simultaneously.
+ The @kbd{I} key adds a new (regular) file to the archive. The file
+is initially empty, but can readily be edited using the commands
+above. The command inserts the new file on the line above the current
+one, so that using it on the topmost line of the Tar buffer makes the
+new file the first one in the archive, and using it at the end of the
+buffer makes it the last one.
+
@kbd{d} marks a file for deletion when you later use @kbd{x}, and
@kbd{u} unmarks a file, as in Dired. @kbd{C} copies a file from the
archive to disk and @kbd{R} renames a file within the archive.
--- a/lisp/tar-mode.el 2014-08-28 19:18:24 +0000
+++ b/lisp/tar-mode.el 2014-12-07 17:44:40 +0000
@@ -50,9 +50,6 @@
;;
;; o chmod should understand "a+x,og-w".
;;
-;; o It's not possible to add a NEW file to a tar archive; not that
-;; important, but still...
-;;
;; o The code is less efficient that it could be - in a lot of places, I
;; pull a 512-character string out of the buffer and parse it, when I could
;; be parsing it in place, not garbaging a string. Should redo that.
@@ -369,6 +366,80 @@
string)
(tar-parse-octal-integer string))
+(defun tar-new-regular-file-header (filename &optional size time)
+ "Return a Tar header for a regular file.
+The header will lack a proper checksum; use `tar-header-block-checksum'
+to compute one, or request `tar-header-serialize' to do that.
+
+Other tar-mode facilities may also require the data-start header
+field to be set to a valid value.
+
+If SIZE is not given or nil, it defaults to 0.
+If TIME is not given or nil, assume now."
+ (make-tar-header
+ nil
+ filename
+ #o644 0 0 (or size 0)
+ (or time (current-time))
+ nil ; checksum
+ nil nil
+ nil nil nil nil nil))
+
+(defun tar--pad-to (pos)
+ (make-string (+ pos (- (point)) (point-min)) 0))
+
+(defun tar--put-at (pos val &optional fmt mask)
+ (when val
+ (insert (tar--pad-to pos)
+ (if fmt
+ (format fmt (if mask (logand mask val) val))
+ val))))
+
+(defun tar-header-serialize (header &optional update-checksum)
+ "Return the serialization of a Tar HEADER as a string.
+This function calls `tar-header-block-check-checksum' to ensure the
+checksum is correct.
+
+If UPDATE-CHECKSUM is non-nil, update HEADER with the newly-computed
+checksum before doing the check."
+ (with-temp-buffer
+ (set-buffer-multibyte nil)
+ (let ((encoded-name
+ (encode-coding-string (tar-header-name header)
+ tar-file-name-coding-system)))
+ (unless (< (length encoded-name) 99)
+ ;; FIXME: Implement it.
+ (error "Long file name support is not implemented"))
+ (insert encoded-name))
+ (tar--put-at tar-mode-offset (tar-header-mode header) "%6o\0 " #o777777)
+ (tar--put-at tar-uid-offset (tar-header-uid header) "%6o\0 " #o777777)
+ (tar--put-at tar-gid-offset (tar-header-gid header) "%6o\0 " #o777777)
+ (tar--put-at tar-size-offset (tar-header-size header) "%11o ")
+ (insert (tar--pad-to tar-time-offset)
+ (tar-octal-time (tar-header-date header))
+ " ")
+ ;; Omit tar-header-checksum (tar-chk-offset) for now.
+ (tar--put-at tar-linkp-offset (tar-header-link-type header))
+ (tar--put-at tar-link-offset (tar-header-link-name header))
+ (when (tar-header-magic header)
+ (tar--put-at tar-magic-offset (tar-header-magic header))
+ (tar--put-at tar-uname-offset (tar-header-uname header))
+ (tar--put-at tar-gname-offset (tar-header-gname header))
+ (tar--put-at tar-dmaj-offset (tar-header-dmaj header) "%7o\0" #o7777777)
+ (tar--put-at tar-dmin-offset (tar-header-dmin header) "%7o\0" #o7777777))
+ (tar--put-at 512 "")
+ (let ((ck (tar-header-block-checksum (buffer-string))))
+ (goto-char (+ (point-min) tar-chk-offset))
+ (delete-char 8)
+ (insert (format "%6o\0 " ck))
+ (when update-checksum
+ (setf (tar-header-checksum header) ck))
+ (tar-header-block-check-checksum (buffer-string)
+ (tar-header-checksum header)
+ (tar-header-name header)))
+ ;; .
+ (buffer-string)))
+
(defun tar-header-block-checksum (string)
"Compute and return a tar-acceptable checksum for this block."
@@ -547,6 +618,7 @@ defvar tar-mode-map
(define-key map "p" 'tar-previous-line)
(define-key map "\^P" 'tar-previous-line)
(define-key map [up] 'tar-previous-line)
+ (define-key map "I" 'tar-new-entry)
(define-key map "R" 'tar-rename-entry)
(define-key map "u" 'tar-unflag)
(define-key map "v" 'tar-view)
@@ -731,10 +803,14 @@
(interactive "p")
(tar-next-line (- arg)))
+(defun tar-current-position ()
+ "Return the `tar-parse-info' index for the current line."
+ (count-lines (point-min) (line-beginning-position)))
+
(defun tar-current-descriptor (&optional noerror)
"Return the tar-descriptor of the current line, or signals an error."
;; I wish lines had plists, like in ZMACS...
- (or (nth (count-lines (point-min) (line-beginning-position))
+ (or (nth (tar-current-position)
tar-parse-info)
(if noerror
nil
@@ -948,6 +1024,37 @@
(write-region start end to-file nil nil nil t)))
(message "Copied tar entry %s to %s" name to-file)))
+(defun tar-new-entry (filename &optional index)
+ "Insert a new empty regular file before point."
+ (interactive "*sFile name: ")
+ (let* ((buffer (current-buffer))
+ (index (or index (tar-current-position)))
+ (d-list (and (not (zerop index))
+ (nthcdr (+ -1 index) tar-parse-info)))
+ (pos (if d-list
+ (tar-header-data-end (car d-list))
+ (point-min)))
+ (new-descriptor
+ (tar-new-regular-file-header filename)))
+ ;; Update the data buffer; fill the missing descriptor fields.
+ (with-current-buffer tar-data-buffer
+ (goto-char pos)
+ (insert (tar-header-serialize new-descriptor t))
+ (setf (tar-header-data-start new-descriptor)
+ (copy-marker (point) nil)))
+ ;; Update tar-parse-info.
+ (if d-list
+ (setcdr d-list (cons new-descriptor (cdr d-list)))
+ (setq tar-parse-info (cons new-descriptor tar-parse-info)))
+ ;; Update the listing buffer.
+ (save-excursion
+ (goto-char (point-min))
+ (forward-line index)
+ (let ((inhibit-read-only t))
+ (insert (tar-header-block-summarize new-descriptor) ?\n)))
+ ;; .
+ index))
+
(defun tar-flag-deleted (p &optional unflag)
"In Tar mode, mark this sub-file to be deleted from the tar file.
With a prefix argument, mark that many files."
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.