From: Markus <markus@...>
Date: 2004-10-27T23:25:59+09:00
Subject: Re: Errors in line numbers reported?

On Wed, 2004-10-27 at 07:12, ts wrote:
> >>>>> "M" == Markus  <markus@reality.com> writes:
> 
> M>       * I came up with a patch (see earlier in this thread) that neatly
> M>         handled line numbers greater than 8k; I added a long to the
> M>         header.  (Previously, they had been packed into bits in the
> M>         flags field).  This maintained double-word-alignment, but grew
> M>         the structure by four bytes.
> 
>  Well, if you don't want to look at rb_newobj() then look at
>  rb_node_newnode() in parse.y

     *laugh*  I've looked at both, but looking may not be seeing.  I
added the following to rb_node_newnode():

    n->u3_state = a0;

which I subsequently changed to:

    nd_set_state(n,a0);

and then:

    if (a0) nd_set_state(n,3); else nd_set_state(n,0); /* kludge */

as I nailed down the values needed for status.  That part seems to
work.  But I am still having trouble running "make test" and it appears
that something else is coulouring outside the lines.

-- Markus

P.S. If you are interested, I could make up a simple buggy-patch to show
you how far I've gotten.