AW: AW: bugfix KHTMLPart::started(KIO::Job *job)

[Grimm, Cornelius]

Hi,

I have another bug fix suggestion...

In khtml_part.cpp replace the last line of slotLoaderRequestDone():
<  checkCompleted();
with:
>  if(d->m_doc && d->m_doc->docLoader() == dl)
>    checkCompleted();

That would fix the problem, that sometimes the completed() signal is sent
before the page is completely loaded.

For more information please see my last mail (attached).

Cornelius


> -----Ursprüngliche Nachricht-----
> Von: Grimm, Cornelius 
> Gesendet: Dienstag, 2. September 2003 11:17
> An: 'kfm-devel at kde.org'
> Betreff: AW: AW: bugfix KHTMLPart::started(KIO::Job *job)
> 
> 
> Hi,
> 
> > > Now, if "slotLoaderRequestDone()" get's called _before_ 
> the document
> > > exists, checkComplete does the completion stuff. And the 
> document is
> > > created in begin(), which is called when the first piece of 
> > data arrives.
> > > This might happen later than the call of 
> "slotLoaderRequestDone()".
> > 
> > Do you use a user stylesheet? Otherwise I cannot imagine why 
> > slotLoaderRequestDone would be called before the document 
> > exists, but it 
> > would be interesting to find out why and then fix it. 
> 
> I found out why.
> 
> In the init() function of KHTMLPart we connect to the 
> requestDone() signal
> of the global loader object returned by khtml::Cache::loader().
> 
> So the slot slotLoaderRequestDone() of KHTMLPart gets called 
> whenever any
> other request is complete, even in completely different 
> KHTMLPart windows
> within this application.
> 
> The implementation of slotLoaderRequestDone() should only call
> checkCompleted() if the khtml::DocLoader belongs to this 
> KHTMLPart, but I
> don't know how to check this (because the document which 
> knows about it's
> loader might be non-existent when the request failed, or not?).
> 
> Cornelius
>