Re: EGexecute: don't assume buffer ends in a newline

[Paolo Bonzini]

On 05/24/2010 12:54 PM, Bruno Haible wrote:
> Hi Paolo,
>
> > >          /* If we've made it to this point, this means DFA has seen
> > >             a probable match, and we need to run it through Regex. */
> > > -      best_match = end;
> > > +      best_match = end + 1;
> >
> > Shouldn't this be
> >
> >     best_match = end + (end<  buflim);
> >
> > ?
>
> No. You see that and the end of the search loop, there is a
>      if (best_match<  end)
> that I turned into
>      if (best_match<= end)
> This statement is meant to detect whether a match has been found at all.
> So, the initial value of best_match has to be greater than 'end' in all cases.

Right.  That's violating ISO C's rule for validity of pointers though. 
In practice I don't think it matters, but I'd rather get a second 
opinion on this patch.

> > Otherwise, the patch seems okay.  Are you going to patch Fexecute too?
>
> You fixed Fexecute already. With msggrep and its use of buffers that don't
> end in a newline, I did not detect a bug in Fexecute. Only in EGexecute.

Ah, okay, I didn't understand that.

Paolo