Allan Rae <[EMAIL PROTECTED]> writes:
| On Wed, 5 Jun 2002, Allan Rae wrote:
>
>> It's not quite as simple as errors after a newline not being removed.
>> Some are, but those in the overly long testcase aren't. I will
>> investigate this weekend.
>
| I've figured it out. This change:
>
| 2002-03-13 Lars Gullik Bjønnes <[EMAIL PROTECTED]>
>
| * paragraph.C: change several algorithm to be more appripriate for
| the problem domain. This is lookip in FontList and in the InsetList.
Hmm, ... bad boy...
| modified Paragraph::InsetIterator(pos) so that it no longer returned
| an inset_iterator with (it.pos >= pos) but now only returns one if
| (it.pos == pos). The net result being that removeAutoInsets() can't
| iterate through the inset list deleting auto-insets as it goes by
| simply calling InsetIterator(pos) -- unless the insets are
| consecutive like they very often are.
>
| Three possible solutions:
| 1. Fix InsetIterator(pos)
| 2. Add nextInsetIterator(pos) with these semantics
| 3. Workaround InsetIterator() in removeAutoInsets()
>
| I picked option 2 since there may be other places that would also
| expect the next nearest inset to be returned rather than
| insetlist.end() when an inset doesn't exist at pos. Of course that
| will only come as paragraph.C is cleaned up to use these helper
| functions instead of implementing it locally in every second function.
>
| Any particular reason why InsetIterator() is capitalised? Is it
| because someone was thinking it looks/acts like a constructor?
| André? (cvs annotate blames you for the name)
>
| Anyway, I've attached a patch against 1.2.1cvs -- it should apply and
| work with 1.3.0cvs but I can't compile that tree at present
| (automake-1.4 not acceptable and upgrading is a low priority).
I wonder why I am able to compile with automake 1.4...
--
Lgb
