🚀 go-pugleaf

In article <1027@sauron.Columbia.NCR.COM> wescott@sauron.Columbia.NCR.COM (Mike Wescott) writes:
>In a number of places in its source, vi uses isdigit() (see ctype(3))
>to check a character (encoded in an int or short) it just received.  No
>problem, except that occasionally, like just after a SIGINT, the character
>
>There may be other isxxxx() calls that have similar problems.
>
>I don't have a fix yet.  We only discovered it while playing with a
>shared library version of ctype.  And there is little probability
>
>Did isdigit() once have range checking built in?  Or was it a version
>called _isdigit()?
>

I've had the bug fix for this since Sept, sorry I haven't posted it yet.
Here it is.  This will get me off by butt and get the rest of the bug
fixes that I've been sitting on posted.

From gww Tue Sep  8 18:16:28 1987
To: /RCS/Bugs

Subject: Vi core dumps on SIGINT input. +Fix
Index:	ucb/ex/{ex_vget.c, ex_vmain.c, ex_voper.c} 4.3BSD +Fix

Description:
	When vi receives the a SIGINT, it can queue it as the input character
	ATTN (value -2).  This character is returned by getkey() (peekkey()).
	In some cases, vi asks if the character received is an xxx with
	``isxxx''.  The ``isxxx'' macros index the array _ctype_[] to
	determine the type of the character.  If _ctype_ begins on a page
	boundary and the previous page is not a valid page, a segment fault
	will occur.
Repeat-By:
	vi foo
	<user types INTERRUPT character>
	When _ctype_ begins on a page boundary and the previous page is not
	valid.
Fix:
	Guard against checking indexing _ctype_ when the character may be
	ATTN.
	The attached code solves this problem at Elxsi.

Gary..
{ucbvax!sun,lll-lcc!lll-tis,amdahl!altos86,bridge2}!elxsi!gww
--------- cut --------- snip --------- :.,$w diff -------------
Index: ucb/ex/ex_vget.c
*** /tmp/,RCSt1001661	Tue Sep  8 17:32:37 1987
--- ex_vget.c	Fri May  8 17:08:01 1987
***************
*** 1,5 ****
--- 1,9 ----
  /*
   * $Log:	ex_vget.c,v $
+  * Revision 1.2  87/05/08  17:06:28  gww
+  * Check for ATTN (-2) before using isxxxx macro so as not to negatively
+  * index _ctype_[] and possibly get a segment fault.
+  * 
   * Revision 1.1  86/12/23  18:16:10  gww
   * Initial revision
   * 
***************
*** 11,17 ****
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_vget.c,v 1.1 86/12/23 18:16:10 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_vget.c	6.8 (Berkeley) 6/7/85";
  #endif not lint
  
--- 15,21 ----
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_vget.c,v 1.2 87/05/08 17:06:28 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_vget.c	6.8 (Berkeley) 6/7/85";
  #endif not lint
  
***************
*** 624,630 ****
  	cnt = 0;
  	for (;;) {
  		c = getkey();
! 		if (!isdigit(c))
  			break;
  		cnt *= 10, cnt += c - '0';
  	}
--- 628,634 ----
  	cnt = 0;
  	for (;;) {
  		c = getkey();
! 		if (c == ATTN || !isdigit(c))
  			break;
  		cnt *= 10, cnt += c - '0';
  	}

Index: ucb/ex/ex_vmain.c
*** /tmp/,RCSt1001666	Tue Sep  8 17:32:57 1987
--- ex_vmain.c	Fri May  8 17:08:07 1987
***************
*** 1,5 ****
--- 1,9 ----
  /*
   * $Log:	ex_vmain.c,v $
+  * Revision 1.2  87/05/08  17:08:01  gww
+  * Check for ATTN (-2) before using isxxxx macro so as not to negatively
+  * index _ctype_[] and possibly get a segment fault.
+  * 
   * Revision 1.1  86/12/23  18:16:13  gww
   * Initial revision
   * 
***************
*** 11,17 ****
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_vmain.c,v 1.1 86/12/23 18:16:13 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_vmain.c	7.7 (Berkeley) 6/7/85";
  #endif not lint
  
--- 15,21 ----
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_vmain.c,v 1.2 87/05/08 17:08:01 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_vmain.c	7.7 (Berkeley) 6/7/85";
  #endif not lint
  
***************
*** 112,118 ****
  			if (trace)
  				fprintf(trace, "pc=%c",peekkey());
  #endif
! 			if (isdigit(peekkey()) && peekkey() != '0') {
  				hadcnt = 1;
  				cnt = vgetcnt();
  				forbid (cnt <= 0);
--- 116,124 ----
  			if (trace)
  				fprintf(trace, "pc=%c",peekkey());
  #endif
! 			if (peekkey() != ATTN && 
! 			    isdigit(peekkey()) && 
! 			    peekkey() != '0') {
  				hadcnt = 1;
  				cnt = vgetcnt();
  				forbid (cnt <= 0);
***************
*** 126,132 ****
  			 * an 'empty' named buffer spec in the routine
  			 * kshift (see ex_temp.c).
  			 */
! 			forbid (c == '0' || !isalpha(c) && !isdigit(c));
  			vreg = c;
  		}
  reread:
--- 132,139 ----
  			 * an 'empty' named buffer spec in the routine
  			 * kshift (see ex_temp.c).
  			 */
! 			forbid (c == '0' ||
! 				c != ATTN && !isalpha(c) && !isdigit(c));
  			vreg = c;
  		}
  reread:
***************
*** 158,164 ****
  			 * to go back to the "for" to interpret it. Likewise
  			 * for a buffer name.
  			 */
! 			if ((isdigit(c) && c!='0') || c == '"') {
  				ungetkey(c);
  				goto looptop;
  			}
--- 165,171 ----
  			 * to go back to the "for" to interpret it. Likewise
  			 * for a buffer name.
  			 */
! 			if ((c != ATTN && isdigit(c) && c!='0') || c == '"') {
  				ungetkey(c);
  				goto looptop;
  			}

Index: ucb/ex/ex_voper.c
*** /tmp/,RCSt1001671	Tue Sep  8 17:34:22 1987
--- ex_voper.c	Fri May  8 17:08:15 1987
***************
*** 1,5 ****
--- 1,9 ----
  /*
   * $Log:	ex_voper.c,v $
+  * Revision 1.2  87/05/08  17:08:08  gww
+  * Check for ATTN (-2) before using isxxxx macro so as not to negatively
+  * index _ctype_[] and possibly get a segment fault.
+  * 
   * Revision 1.1  86/12/23  18:16:15  gww
   * Initial revision
   * 
***************
*** 11,17 ****
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_voper.c,v 1.1 86/12/23 18:16:15 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_voper.c	7.4 (Berkeley) 6/7/85";
  #endif not lint
  
--- 15,21 ----
   */
  
  #ifndef lint
! static char *ERcsId = "$Header: ex_voper.c,v 1.2 87/05/08 17:08:08 gww Exp $ ENIX BSD";
  static char *sccsid = "@(#)ex_voper.c	7.4 (Berkeley) 6/7/85";
  #endif not lint
  
***************
*** 133,139 ****
  	 * Had an operator, so accept another count.
  	 * Multiply counts together.
  	 */
! 	if (isdigit(peekkey()) && peekkey() != '0') {
  		cnt *= vgetcnt();
  		Xcnt = cnt;
  		forbid (cnt <= 0);
--- 137,143 ----
  	 * Had an operator, so accept another count.
  	 * Multiply counts together.
  	 */
! 	if (peekkey() != ATTN && isdigit(peekkey()) && peekkey() != '0') {
  		cnt *= vgetcnt();
  		Xcnt = cnt;
  		forbid (cnt <= 0);

The is...() macros are only valid to test the type of ASCII values;
that's what the isascii() macro is for.

The correct usage is:
	if(isascii(c) && isdigit(c)) ...

and that's how it should be fixed in vi.
(See K&R p.127 or H&S p. 291).

-- 
	Amos Shapir				My other CPU is a NS32532
National Semiconductor 7C/266  1135 Kern st. Sunnyvale (408) 721-8161
amos@nsc.com till March 1, 88; Then back to amos%taux01@nsc.com 

In article <1027@sauron.Columbia.NCR.COM> wescott@sauron.Columbia.NCR.COM (Mike Wescott) writes:
>[vi uses isdigit() on an out-of-band value]
>Did isdigit() once have range checking built in?  Or was it a version
>called _isdigit()?

Not that I know of.  But a simple fix is
  #define isdigit(c) ((unsigned)((c)-'0') < (unsigned)10)
which will work if the digits are contiguous.  (This is true of every
character set I'm familiar with, and required by ANSI C.)

(I prefer to use such macros rather than <ctype.h>, because many systems still
implement the latter with a 128-character range, and I like my software to
work even when you hand it 8-bit data.)

Karl W. Z. Heuer (ima!haddock!karl or karl@haddock.isc.com), The Walking Lint