Re: BUG REPORT tin 1.9.5 release 20091201 ("Rieclachan") [UNIX]

On 09.12.2009, at 17:30, Urs Janßen wrote:

> tin crashes when viewing  
> <[email protected]>
> in news.software.readers while building the page header (inside the  
> References).

[...]

> the headers of mentioned article:

> References: <[email protected]>
> 	<[email protected]>
> 	<[email protected]>
> 	<[email protected]>
> 	<[email protected]>

In cook.c:(w)expand_ctrl_chars() we resize the buffer too late. At  
least one expanded TAB (+ \0) should fit into the remaining space.

Another issue: If an header exceeds LEN (1024) after expanding, only  
the first 1024 chars are visible. cook.c:cook_article() uses LEN in  
put_cooked(). The patch below changes this too (use the real length).

diff -urp tin-1.9.5/src/cook.c tin-1.9.5_r1/src/cook.c
--- tin-1.9.5/src/cook.c	2009-12-01 15:15:28.000000000 +0100
+++ tin-1.9.5_r1/src/cook.c	2009-12-10 19:57:34.000000000 +0100
@@ -111,7 +111,7 @@ expand_ctrl_chars(

  	c = (unsigned char *) *line;
  	while (*c) {
-		if (i > curr_len - 3) {
+		if (i > curr_len - (lcook_width + 1)) {
  			curr_len <<= 1;
  			buf = my_realloc(buf, curr_len);
  		}
@@ -158,14 +158,15 @@ wexpand_ctrl_chars(

  	wc = *wline;
  	while (*wc) {
-		if (i > cur_len - 3) {
+		if (i > cur_len - (lcook_width + 1)) {
  			cur_len <<= 1;
  			wbuf = my_realloc(wbuf, cur_len * sizeof(wchar_t));
  		}
@@ -848,7 +849,7 @@ cook_article(
  #endif /* MULTIBYTE_ABLE && !NO_LOCALE */
  			header_put = TRUE;
  			expand_ctrl_chars(&l, &i, tabwidth);
-			put_cooked(LEN, wrap_lines, C_HEADER, "%s", l);
+			put_cooked(i, wrap_lines, C_HEADER, "%s", l);
  			free(l);
  		}
  	}

Last but not least a question: in cook.c:(w)expand_ctrl_chars() we  
use the following code to compute j:

| if (*wc == '\n')
| 	ln = i + 1;
| if (*wc == '\t') {		/* expand_tabs */
| 	j = i + lcook_width - ((i - ln) % lcook_width);
| 	for (; i < j; i++)
| 		wbuf[i] = ' ';

This gives wrong results when a line contains several TABs (but no  
newlines).

I think it is sufficient to use only j = i + lcook_width here (no  
need for ln), but I'm not entirely shure whether this breaks anything  
else or not.

Dennis