Buffer corruption with jka-compr.el and revert-buffer

Buffer corruption with jka-compr.el and revert-buffer

[Martin Schwenke]

In GNU Emacs 21.3.2 and XEmacs 21.4.12 (and lots of earlier versions
of each, apparently, such as GNU Emacs 20.7), put the following code
into jka-compr-revert-bug.el:

--------8<---------8<-------- CUT HERE --------8<---------8<--------
(shell-command "echo 'aaa' | gzip > /tmp/aa.gz")
(shell-command "echo 'b' | gzip > /tmp/b.gz")
(auto-compression-mode 1)
(find-file "/tmp/aa.gz")
(shell-command "cp /tmp/b.gz /tmp/aa.gz")
(revert-buffer t t)
--------8<---------8<-------- CUT HERE --------8<---------8<--------

Now do:

  (x)emacs -q --no-site-file -l jka-compr-revert-bug.el

You probably have a buffer that looks like

--------8<---------8<-------- CUT HERE --------8<---------8<--------
b
a
--------8<---------8<-------- CUT HERE --------8<---------8<--------

Yikes!  It should just contain a 'b'!

This only seems to happen if the original file is at least 2
characters longer than the new file.  The underlying file is not
changed but, when you edit and save, the corruption obviously makes
its way into the saved file.

I've had a look at revert-buffer and nothing jumps out.  I've also had
a look at jka-compr.el and I've run away screaming...  I suspect it
would be a lot simpler without all that coding system stuff...  :-)

peace & happiness,
martin

Re: Buffer corruption with jka-compr.el and revert-buffer

[Stephen J. Turnbull]

Thanks for your report, Martin.

>>>>> "Martin" == Martin Schwenke <martin@meltin.net> writes:

    Martin> I've had a look at revert-buffer and nothing jumps out.
    Martin> I've also had a look at jka-compr.el and I've run away
    Martin> screaming...  I suspect it would be a lot simpler without
    Martin> all that coding system stuff...  :-)

The REPLACE semantics of j-c-insert-file-contents were simply wrong.

This fixes that, and a couple of minor bugs like the truncated
argument list and the missing docstring.  It's not really ready for
prime time; see the #### comments.  However, it's strictly better than
the existing code in XEmacs, and jka-compr will probably be obsolete
in XEmacs 21.5.

Caveat Emacsers.  I have no idea whether this is correct for GNU Emacs.

Index: xemacs-packages/os-utils/ChangeLog
===================================================================
RCS file: /pack/xemacscvs/XEmacs/packages/xemacs-packages/os-utils/ChangeLog,v
retrieving revision 1.38
diff -u -U0 -r1.38 ChangeLog
--- xemacs-packages/os-utils/ChangeLog	21 Mar 2003 05:51:13 -0000	1.38
+++ xemacs-packages/os-utils/ChangeLog	26 May 2003 09:24:01 -0000
@@ -0,0 +1,5 @@
+2003-05-26  Stephen J. Turnbull  <stephen@xemacs.org>
+
+	* jka-compr.el (jka-compr-insert-file-contents): New docstring.
+	Implement REPLACE argument correctly.
+

Index: xemacs-packages/os-utils/jka-compr.el
===================================================================
RCS file: /pack/xemacscvs/XEmacs/packages/xemacs-packages/os-utils/jka-compr.el,v
retrieving revision 1.8
diff -u -r1.8 jka-compr.el
--- xemacs-packages/os-utils/jka-compr.el	27 Mar 2000 13:11:15 -0000	1.8
+++ xemacs-packages/os-utils/jka-compr.el	26 May 2003 09:24:08 -0000
@@ -517,7 +517,31 @@
 					  )))))
 
 
-(defun jka-compr-insert-file-contents (file &optional visit beg end replace)
+;; #### Should error or warn if the ignore-me arguments are supplied.
+(defun jka-compr-insert-file-contents (file &optional visit beg end replace
+				       ignore-me ignore-me-too)
+  "Insert contents of FILE in current buffer after point.
+Returns list of absolute file name and length of data inserted.
+
+If second argument VISIT is non-nil, the buffer's visited filename
+and last save file modtime are set, and it is marked unmodified.
+If visiting and the file does not exist, visiting is completed
+before the error is signaled.
+
+The optional third and fourth arguments START and END
+specify what portion of the file to insert.
+If VISIT is non-nil, START and END must be nil.
+If optional fifth argument REPLACE is non-nil,
+it means replace the current buffer contents (in the accessible portion)
+with the file contents.  This is better than simply deleting and inserting
+the whole thing because (1) it preserves some marker positions
+and (2) it puts less data in the undo list.  (#### Is (2) true?)
+
+Coding system arguments are completely ignored (see documentation of the
+6th and 7th arguments to `insert-file-contents').  The coding system is set
+to 'undecided unless `coding-system-for-read' is bound.  WARNING: That means
+this function is badly broken under Mule.  Use jka-compr at your own risk
+in Mule Emacsen."
   (barf-if-buffer-read-only)
 
   (and (or beg end)
@@ -571,7 +594,7 @@
 
 		    (progn
 		      (if replace
-			  (goto-char (point-min)))
+			  (delete-region (point-min) (point-max)))
 		      (setq start (point))
 		      (if (or beg end)
 			  (jka-compr-partial-uncompress uncompress-program
@@ -595,12 +618,6 @@
 						  t
 						  nil
 						  uncompress-args)))
-		      (setq size (- (point) start))
-		      (if replace
-			  (let* ((del-beg (point))
-				 (del-end (+ del-beg size)))
-			    (delete-region del-beg
-					   (min del-end (point-max)))))
 		      (goto-char start))
 		  (error
 		   (if (and (eq (car error-code) 'file-error)



-- 
Institute of Policy and Planning Sciences     http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
               Ask not how you can "do" free software business;
              ask what your business can "do for" free software.

Re: Buffer corruption with jka-compr.el and revert-buffer

[Martin Schwenke]

Hi Stephen,

>>>>> "Stephen" == Stephen J Turnbull <stephen@xemacs.org> writes:

    Stephen> Thanks for your report, Martin.

Any time.  Thanks for the fix!

    Stephen> The REPLACE semantics of j-c-insert-file-contents were
    Stephen> simply wrong.

    Stephen> This fixes that, and a couple of minor bugs [...]

    Stephen> Caveat Emacsers.  I have no idea whether this is correct
    Stephen> for GNU Emacs.

Party time Emacsers: it does seem to fix the problem in GNU Emacs
21.3.

peace & happiness,
martin

Re: Buffer corruption with jka-compr.el and revert-buffer

[Martin Schwenke]

>>>>> "Martin" == Martin Schwenke <martin@meltin.net> writes:

    Stephen> The REPLACE semantics of j-c-insert-file-contents were
    Stephen> simply wrong.

    Stephen> This fixes that, and a couple of minor bugs [...]

    Stephen> Caveat Emacsers.  I have no idea whether this is correct
    Stephen> for GNU Emacs.

    Martin> Party time Emacsers: it does seem to fix the problem in
    Martin> GNU Emacs 21.3.

Yikes, but not when the revert-buffer happens interactively.  With the
fix, even when you tell revert-buffer not to prompt (via
(revert-buffer t t) or by setting revert-without-query), it prompts
but refuses to accept 'r' as an answer.

I'll play with this and try to figure out what is happening...

peace & happiness,
martin