Re: subtle byte-compiler bug

Re: subtle byte-compiler bug

[David PONCE]

Hi,

Oliver Scholz wrote:
> David Ponce (CCed here) suggested that this is a bug in the
> byte-compiler. He also noted that this bug is not in 21.3.
[...]

I finally got a little time to investigate on that problem and
discovered that the bug is not in the byte-compiler but in the
`prin1' primitive.

Sometimes `prin1' print incorrect uninterned symbol references,
that can result in writing invalid byte-compiled code.

Here is a small example of code that illustrates the bug:

(let* ((n 512)
       (print-escape-newlines t)
       (print-length nil)
       (print-level nil)
       (print-quoted t)
       (print-gensym t)
       (v (make-vector n (list 1 2 3)))
       (tmp1 (make-symbol "defconst-tmp-var"))
       (tmp2 (make-symbol "defconst-tmp-var"))
       )
  (aset v 3 `(lambda (,tmp1) (defconst v1  ,tmp1)))
  (aset v 8 `(lambda (,tmp2) (defconst v2 ,tmp2)))
  (prin1 v))

When I eval it in the *scratch* buffer, `prin1' print something like
the following, which is obviously incorrect :-(

[(1 2 3) (1 2 3) (1 2 3)
 (lambda (#:defconst-tmp-var) (defconst v1 #:defconst-tmp-var))
 (1 2 3) (1 2 3) (1 2 3) (1 2 3)
 (lambda (#:defconst-tmp-var) (defconst v2 #:defconst-tmp-var))
 (1 2 3) (1 2 3) (1 2 3) (1 2 3) (1 2 3) (1 2 3) ...]

However, If I simply change (n 512) to (n 511) in the `let'
expression, the printed result is then correct:

[(1 2 3) (1 2 3) (1 2 3)
 (lambda (#1=#:defconst-tmp-var) (defconst v1 #1#))
 (1 2 3) (1 2 3) (1 2 3) (1 2 3)
 (lambda (#2=#:defconst-tmp-var) (defconst v2 #2#))
 (1 2 3) (1 2 3) (1 2 3) (1 2 3) (1 2 3) (1 2 3) ...]

I am running:

GNU Emacs 21.3.50.1 (i686-pc-linux-gnu, GTK+ Version 2.2.4)
 of 2004-01-07
configured using `configure '--prefix=/home/ponce' '--with-x-toolkit=gtk''

Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: en_US.UTF-8
  locale-coding-system: utf-8
  default-enable-multibyte-characters: t

Hope this will help.

Sincerely,
David

Re: subtle byte-compiler bug

[Andreas Schwab]

David PONCE <david.ponce@wanadoo.fr> writes:

> I finally got a little time to investigate on that problem and
> discovered that the bug is not in the byte-compiler but in the
> `prin1' primitive.
>
> Sometimes `prin1' print incorrect uninterned symbol references,
> that can result in writing invalid byte-compiled code.

Thanks for the testcase.  I have installed this change to fix the bug:

	* print.c (print_preprocess) <case Lisp_Vectorlike>: Only mask
	size if PSEUDOVECTOR_FLAG is set.

Index: print.c
===================================================================
RCS file: /cvsroot/emacs/emacs/src/print.c,v
retrieving revision 1.196
retrieving revision 1.197
diff -u -a -p -a -u -p -r1.196 -r1.197
--- print.c	31 Dec 2003 00:24:28 -0000	1.196
+++ print.c	8 Jan 2004 00:15:52 -0000	1.197
@@ -1342,7 +1342,9 @@ print_preprocess (obj)
 	  goto loop;
 
 	case Lisp_Vectorlike:
-	  size = XVECTOR (obj)->size & PSEUDOVECTOR_SIZE_MASK;
+	  size = XVECTOR (obj)->size;
+	  if (size & PSEUDOVECTOR_FLAG)
+	    size &= PSEUDOVECTOR_SIZE_MASK;
 	  for (i = 0; i < size; i++)
 	    print_preprocess (XVECTOR (obj)->contents[i]);
 	  break;

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: subtle byte-compiler bug

[Richard Stallman]

Thanks for fixing this, and thanks very much David for tracking it down.

Re: subtle byte-compiler bug

[David PONCE]

Hi Andreas,

> Thanks for the testcase.  I have installed this change to fix the bug:
> 
> 	* print.c (print_preprocess) <case Lisp_Vectorlike>: Only mask
> 	size if PSEUDOVECTOR_FLAG is set.

It works like a charm now :-)
Thanks!

David