On Mon, Aug 14, 2006 at 05:21:49PM +0200, Jean-Marc Lasgouttes wrote: > > The following patch fixes this bug for me. However I have questions > about whether I did it right. > > There are 3 mostly independent parts in the patch: > > - in MathScriptInset::notifyCursorLeave, there were no recordUndoInset > calls. This fix seems safe and obvious to me
Worst case scenario is that an additional C-z is needed to undo stuff. Could be mildly annoying if it is (not sure it is...) but would certainly acceptable to fix a crash. > - when doing record undo, the math cell is stored as a string. > However, MathScriptInset::write omits empty scripts. I changed this > in the script, but I do not know whether there are bad side effects. Saving a document is changed. Try a round trip with empty cells. The potential of damage is low in any case as empty scripts can be read back correctly. Might be even a good move in general. > An alternative would be to add a keepempty member to WriteStream, > which would be set to true when in asString. Sounds more difficult and fragile. > - the part in undo is not really related to the bug (it could be > omitted): it ensures that the cursor is set correctly after undoing > an recordUndoInset undo. If it works ;-} Andre'
