Re: Xterm Bidi/Shaping/etc Patch

[Karl Koehler <koehler@xxxxxxxxxxxxxx>]

Hello Robert,

I tried to run xterm-150 with and without the patch. 
Without the patch everything seems ok. With the patch xterm
* sucks up huge gobs of memory ( a rise from 3256K to 50M (!) within 
  minutes )
* crashes quickly. 
  I traced down:

  
----
  screen.c::798: 

   #if OPT_WIDE_CHARS
    lefter = (chars[lastind-1] + (WIDEC(lastind-1) << 8));

  In the initiazition phase, chars[-1] is accessed, because lastind is 
  zero. This seems at least strange.
----
  screen.c::1020:
	    /* adjust to redraw all of a widechar if we just wanted 
	       to draw the right hand half */
	    if (iswide(chars[leftcol - 1] | (widec[leftcol -1]<<8)) &&
		(chars[leftcol] | (widec[leftcol]<<8))==HIDDEN_CHAR)

  leftcol can be 0, and is not checked. Access of chars[-1] results.
  This is not good if the first row is concerned :(.
----

  screen.c::776:
    	chars = malloc(screen->max_col+1);
	widec = malloc(screen->max_col+1);

  But anyway, on line 798 there is:

  righter = screen->wide_chars ? (((comb1l[col]|(comb1h[col])<<8) == ZWNJ ? 0 :
				   (chars[col] + (WIDEC(col) << 8)))) : 0;
  
  and a check reveals that
   col > screen->max_col 
  can appear !!! ( the variable col is passed in to 
  ActuallyDraw as parameter ). ?
 
  This results in a crash if xterm is compiled with -O2.
----
  I added a few comments, the whole thing might be shifted by a few lines.

  The C++ -style comments ( // ) are maybe not such a good idea, if xterm 
  is not compiled with a C++-compiler.
  

I which you'd find a good fix for these errors ( I don't know the code 
well enough ), because I'd hate to miss the features :)


thanks,
		
		Karl


-
Linux-UTF8:   i18n of Linux on all levels
Archive:      http://mail.nl.linux.org/lists/