On Tue, Nov 11, 2003 at 12:32:34PM +0100, Alfredo Braunstein wrote: > This an update rationalization: do not call update on the insets, but return > the update flag in the dispatch result (I've had to put some default update > value to true in dispatchresult.h, > Lars please have a look. If that is no > good, then I'll need a DispatchResult(false, FINISHED_RIGHT, update) for > putting update = true). Only collapsables/insettext handled here (tables > are still unoperational). > > For doing this, I've added an LCursor::updatePos to update the cached y > position of the inset. Both the cached y position and this updatePos can be > removed if we switch to absolute document coords.
Because we'd call innerInset()->y() instead? > Still a lot of cleanup to be done, but seems to work reasonably well > already. > > The work for making put_selection_at is tricky: we still have the > chicken-and-egg problem (i.e. andre you were right, but not when > scrolling): > > if the fitCursor call refers to some point in never-explored-land, then we > don't have y information in insets there, then we cannot even do an update > to get it (because we need the y infomation for setting top_y first). > So this is solved in put_selection_at with: > > - setting top_y to the y of the outerPar (that has good info) > - calling update > - calling cursor().updatePos() > - then call fitCursor() > > Comments? Just put this full explanation there and not only '//hack for the chicken and egg problem' So, fine. Andre'
