[9fans] sed line limit

[9fans] sed line limit

[erik quanstrom]

it appears that sed won't handle an input line longer than 8k characters.
yet no diagnostic is printed.  and this is not mentioned in the
man page.

example
	for(i in `{seq 3000 3849}) x = ($x IMG_$i.JPG)
	echo $x | sed 's/IMG_([0-9]+)\.JPG:\1:g'
no error is printed, yet the output is too short and the tail
is corrupted (... 3565 IM)

i added a diagnostic for input line too long.
in trying to test this with
	cat fileof8192is | sed 's/i/j/g'
i found that a single long line can result in at least
1 "sed: Output line too long." per character of the input
line.  i also addressed this problem.  just one output
too long message is printed per line.

the diff would be shorter if i weren't paranoid
that the output buffer could lose its null terminator.
i couldn't provoke that situation so perhaps i just haven't
read the source carefully enough.

- erik


/n/dump/2010/0315/sys/src/cmd/sed.c:161,166 - sed.c:161,167
  Rune	bad;				/* Dummy err ptr reference */
  Rune	*badp = &bad;

+ int	tlwarn;				/* during sub: have warned too long */

  char	CGMES[]	 = 	"%S command garbled: %S";
  char	TMMES[]	 = 	"Too much text: %S";
/n/dump/2010/0315/sys/src/cmd/sed.c:197,202 - sed.c:198,204
  char	*text(char *);
  Rune	*stext(Rune *, Rune *);
  int	ycomp(SedCom *);
+ void	toolong(void);
  char *	trans(int c);
  void	putline(Biobuf *bp, Rune *buf, int n);
  void	ebputc(Biobufhdr*, int);
/n/dump/2010/0315/sys/src/cmd/sed.c:697,705 - sed.c:699,708
  int
  rline(Rune *buf, Rune *end)
  {
- 	long c;
+ 	long c, w;
  	Rune r;

+ 	w = 0;
  	while ((c = getrune()) >= 0) {
  		r = c;
  		if (r == '\\') {
/n/dump/2010/0315/sys/src/cmd/sed.c:714,719 - sed.c:717,726
  		}
  		if (buf <= end)
  			*buf++ = r;
+ 		else if(w == 0){
+ 			fprint(2, "sed: Input line too long.\n");
+ 			w = 1;
+ 		}
  	}
  	*buf = '\0';
  	return -1;
/n/dump/2010/0315/sys/src/cmd/sed.c:1022,1027 - sed.c:1029,1035
  	 * bump to the character after a 0-length match to keep from looping.
  	 */
  	sflag = 1;
+ 	tlwarn = 0;
  	if(ipc->gfl == 0)			/* single substitution */
  		dosub(ipc->rhs);
  	else
/n/dump/2010/0315/sys/src/cmd/sed.c:1065,1083 - sed.c:1073,1096
  				errexit();
  			}
  		}
- 		*sp++ = c;
- 		if (sp >= &genbuf[LBSIZE])
- 			fprint(2, "sed: Output line too long.\n");
+ 		if(sp < &genbuf[LBSIZE]){
+ 			*sp++ = c;
+ 			if (sp >= &genbuf[LBSIZE])
+ 				toolong();
+ 		}
  	}
  	lp = loc2;
  	loc2 = sp - genbuf + linebuf;
  	while (*sp++ = *lp++)
- 		if (sp >= &genbuf[LBSIZE])
- 			fprint(2, "sed: Output line too long.\n");
+ 		if (sp >= &genbuf[LBSIZE]){
+ 			toolong();
+ 			break;
+ 		}
  	lp = linebuf;
  	sp = genbuf;
  	while (*lp++ = *sp++)
- 		;
+ 		if (sp >= &genbuf[LBSIZE])
+ 			break;
  	spend = lp - 1;
  }

/n/dump/2010/0315/sys/src/cmd/sed.c:1086,1097 - sed.c:1099,1120
  {
  	while (l1 < l2) {
  		*sp++ = *l1++;
- 		if (sp >= &genbuf[LBSIZE])
- 			fprint(2, "sed: Output line too long.\n");
+ 		if (sp >= &genbuf[LBSIZE]){
+ 			toolong();
+ 			break;
+ 		}
  	}
  	return sp;
  }

+ void
+ toolong(void)
+ {
+ 	if(tlwarn == 0)
+ 		fprint(2, "sed: Output line too long.\n");
+ 	tlwarn = 1;
+ }
+
  char *
  trans(int c)
  {
/n/dump/2010/0315/sys/src/cmd/sed.c:1408,1414 - sed.c:1431,1437
  Rune *
  gline(Rune *addr)
  {
- 	long c;
+ 	long c, w;
  	Rune *p;
  	static long peekc = 0;

/n/dump/2010/0315/sys/src/cmd/sed.c:1417,1422 - sed.c:1440,1446
  	sflag = 0;
  	lnum++;
  /*	Bflush(&fout);********* dumped 4/30/92 - bobf****/
+ 	w = 0;
  	do {
  		p = addr;
  		for (c = (peekc? peekc: Bgetrune(f)); c >= 0; c = Bgetrune(f)) {
/n/dump/2010/0315/sys/src/cmd/sed.c:1426,1433 - sed.c:1450,1463
  				*p = '\0';
  				return p;
  			}
- 			if (c && p < lbend)
- 				*p++ = c;
+ 			if (c) {
+ 				if (p < lbend)
+ 					*p++ = c;
+ 				else if(w == 0) {
+ 					w = 1;
+ 					fprint(2, "sed: Input line too long.\n");
+ 				}
+ 			}
  		}
  		/* return partial final line, adding implicit newline */
  		if(p != addr) {