Abdelrazak Younes wrote: > Alfredo Braunstein wrote: >> Abdelrazak Younes wrote: >> >>> fixIfBroken). But I don't really understand why you think the signal >>> solution is bad. This is the *only* way to ensure that the inset pointer >>> inside CursorSlice is valid at any time. Even with Alfredo solution you >>> could end up with an invalid pointer if you forgot to call >>> fixIfBroken(). >> >> But you'll still end up with an invalid cursor if you forget to call >> fixIfbroken, > > Right but this case would be detected by an assertion whereas using an > invalid inset pointer will crash LyX very badly.
What assertion? (I mean where in the code). IMO you can end up with an invalid iterator that can blow up anywhere. >> so there's no point in keeping the pointers valids, right? > > I am not sure. The point is to make sure that we know if a CursorSlice > is valid or not. Your idea about comparing the address of the current > inset with the one listed in the insetlist of the enclosing paragraph in > order to verify that the pointer address is valid might work. I am just > saying that the signal approach is safer, that's all. I think not, because the signals are *not* sufficient to determine if the CursorSlice is still valid. Of course I think that all this is not very relevant because the right solution would be to avoid the iterators being invalid in the first place. As a side note, I wished really stable iterators from the very first lyx pet project of mine, errorlists ;-) So it would be a nice thing to have in general. A/
