> normalize() is called only from the LFUN_CUT case of > InsetMathNest::doDispatch(). > It corrects only the pos of the top cursor slice (if needed), AFAICS. > >> Fact is, we should apparently not call normalize() from a non-math >> inset, so there is something wrong there if it happens. > > Is an ASSERT better than a crash than?
Yes, of course. However, normalize() is only called from one place. There it is called after a call to cutSelection(). Cursor::cutSelection either : - does nothing if there is no selection, - calls "cur.pos() = endpos" if the cursor is in Text, or - calls "cur.pos() = cur.lastpos()" or "cur.pos() = 0" if the cursor is in Math. So, cur.pos() seems to be never > cur.lastpos(). If we are in math, we do not delete cells, so cur.idx() cannot be > cur.lastidx(). So.. we can just remove Cursor::normalize() altogether. Vincent
