Buffer::inset_iterator dereferences to InsetBase &.
In turn, this implies that InsetList can hold only non-zero inset
pointers. That's not true:
paragraph_funcs.C:
bool moveItem(Paragraph & from, Paragraph & to,
BufferParams const & params, pos_type i, pos_type j)
{
...
if (from.getInset(i)) {
// the inset is not in a paragraph anymore
tmpinset = from.insetlist.release(i);
}
if (!to.insetAllowed(tmpinset->lyxCode()))
return false;
to.insertInset(j, tmpinset, tmpfont);
}
AFAICS, this is the only instance where InsetList::release is called
--- why not cull it?
Assuming that it's there for a reason, I don't understand why the code
doesn't go on to invoke
from.insetlist.erase(i);
It seems that the list now contains a null inset pointer for no good
reason.
Furthermore, there is a leak in the code above. I think the minimal
change is:
if (from.getInset(i)) {
// the inset is not in a paragraph anymore
tmpinset = from.insetlist.release(i);
+ from.insetlist.erase(i);
}
- if (!to.insetAllowed(tmpinset->lyxCode()))
+ if (!to.insetAllowed(tmpinset->lyxCode())) {
+ delete tmpinset;
return false;
+ }
- to.insertInset(j, tmpinset, tmpfont);
+ if (tmpinset)
+ to.insertInset(j, tmpinset, tmpfont);
Ie, I think it was wrong in THREE places!
Am I going nuts?
--
Angus
