* William Morgan <wmorgan-...@masanjin.net> [Fr Apr 15 19:16:43 +0200 2011]
> Reformatted excerpts from Gregor Hoffleit's message of 2011-04-15:
> > The source of this problem is in whistlepig's tokenizer. I'm hurt by an
> > overflow in posarray->next and posarray->size, which are defined as
> > uint16_t. I was able to fix my problem by defining these to uint32_t:
>
> Thanks! This is totally good. I will admit I got distracted by some
> other tasks and forgot to come back to this. Maybe that's a sign I
> should start using Github's vaunted issue tracker.
>
> What I don't understand is that the posarray stuff is all for token
> offsets, isn't it? So why would the number of lines actually matter?
> Large messages can overflow this but it should be in terms of tokens,
> not lines.
Yep. In fact the number of lines is a pure coincidence.
The mail that got stuck was a logcheck report, which had 35.000 lines
all starting with the current date. I guess whistlepig tried to append
some 35.000 instances of "Feb" to the posarray, which lead to an
infinite loop.
Regards,
Gregor
_______________________________________________
Sup-devel mailing list
Sup-devel@rubyforge.org
http://rubyforge.org/mailman/listinfo/sup-devel
