Re: [pdf-devel] Bug in LZW filter?

[Aleksander Morgado]

> >>> -          if (st->buffer.maxval - st->early_change ==
> >>> st->dict.size) +          if (st->buffer.maxval -
> >>> st->early_change + 2 == st->dict.size) { if
> >>> (!lzw_buffer_inc_bitsize (&st->buffer)) {
> > 
> > I found a very good explanation for a "+1" at this place (a second
> > +1 is of course still missing).
> 
> I have found the second +1
> 
> It is caused by mixing st->dict.size with table entry content:
> ISO-32000 says: "The first output code that is 10 bits long shall be
> the one following the creation of table entry 511"
> (entry 511, not dict.size = 511)
> 
> If we create table entry in function lzw_dict_add with:
>   d->table[d->size++] = *s;
> 
> dict.size is already 511 after the entry of 510. So we have test
> against st->dict.size-1 (or add 1 to maxval)
> 

If that is the case, then it's probably better to test against
dict.size-1. 

As you probably know already, digesting this algorithm is pretty
tough :-) It would be great if you could document all your findings
already in the source code, so that one can follow both the flow and the
comments when reading it.

Cheers!

-- 
Aleksander

signature.asc
Description: This is a digitally signed message part