kdev4 x86_64 build error

[David Nolden]

Am Mittwoch, 11. Juli 2007 01:48:14 schrieb Andreas Pakulat:
> I can say this much: Current kdevelop opens an empty completion box,
> computes some stuff (i.e. "it does something") and then crashes.
Mine doesn't.. Got some backtrace? And better send it private, this is work 
going on, we don't need to spam a hundred people with such things. :)

> Apart from that I already said to David: If the data in the index is a
> number internalId() can also be used and then we don't have to worry
> about sizeof(void*) or casting.
Yes I'll change that at some point if internalId() does what you say it does. 
Currently I cannot commit such a thing because I have other changes waiting 
to commit in the same area.

> Also I still miss an answer (or already forgot it again :) from David
> wether the row fetched from the internal pointer is a different
> row-thing than the row of the model index.
>
> Where are entries into the completion model created? I mean where is
> index() and parent() implemented - in kdevelop or kate?

The row fetched from the internal pointer simply points into an array of 
self-invalidating declarations-pointers stored locally in 
CppCodeCompletionModel. So if I called it "row" I'd better rename it. :)

This is needed to use the self-invalidating du-chain pointers. Before that, 
internalPointer simply returned the declaration-pointer, but nothing can 
guarantee that the declaration is not already deleted. The other option is 
creating a QMetaType for the self-invalidating pointer and attaching it to 
the index. Maybe it'll be better doing it that way in the end.

greetings, David