On Fri, Oct 20, 2006 at 05:08:24PM +0200, Abdelrazak Younes wrote: > Abdelrazak Younes wrote: > >OK, I've found the bloody source of the tabular crash. I don't know what > >this resetPos() method is for but it should be used carefully. > > Here's an alternate patch. Which one do you prefer? > > Martin, can I call this one a "fix"? :-)
This certainly looks more robust. If you're confident that the problem
is indeed about accessing an inset not in the cache...
(I remember there were a lot of such problems around not so long ago...
fixed some myself)
- Martin
> Index: coordcache.h
> ===================================================================
> --- coordcache.h (revision 15403)
> +++ coordcache.h (working copy)
> @@ -54,6 +54,11 @@
> data_.clear();
> }
>
> + bool const empty() const
> + {
> + return data_.empty();
> + }
> +
> void add(T const * thing, int x, int y)
> {
> data_[thing] = Point(x, y);
> Index: insets/insettabular.C
> ===================================================================
> --- insets/insettabular.C (revision 15403)
> +++ insets/insettabular.C (working copy)
> @@ -1236,7 +1236,7 @@
>
> if (&cur.inset() != this) {
> scx_ = 0;
> - } else {
> + } else if (!bv.coordCache().getInsets().empty()) {
> int const X1 = 0;
> int const X2 = maxwidth;
> int const offset = ADD_TO_TABULAR_WIDTH + 2;
pgp57u1hO5ctO.pgp
Description: PGP signature
