Jean-Marc Lasgouttes wrote:
> Juergen> Jean-Marc Lasgouttes wrote: // This is correct and always
> Juergen> safe. // (See Meyer's Effective STL, Item 28.) return
> Juergen> (++rit).base();
>>> Isn't it possible to fix this instead? I agree that the code is
>>> not very clear...
>
> Juergen> I cannot fix it, because I don't understand it (and I
> Juergen> certainly don't own Meyer's book).
>
> You are in the same situation as me, then.
>
> Could a reverse_iterator guru take a quick look? It has to be
> trivial.
Actually, the bug is nothing to do with the internals of searchKeys
which is where the reverse_iterator stuff is used. The bug is, indeed
fixed by Jürgen's patch:
vector<string>::const_iterator start = form_->bibkeys.begin();
int const sel = add_->availableLB->currentItem();
if (sel >= 0 && sel <= int(form_->bibkeys.size()-1))
start += sel;
// Angus' annotation: if sel == 0, above then start ==
// form_->bibkeys.begin()
// Find the NEXT instance...
if (dir == biblio::FORWARD)
start += 1;
else
start -= 1;
Ok, bibkeys.begin() - 1 is 1 before the start. Isn't it undefined
behaviour to do pretty well anything with such an iterator?
--
Angus
