[patch] to bug 88243

[Leo Savernik]

Am Freitag, 11. Februar 2005 19:25 schrieb Koos Vriezen:
> > Because xml/ interferes with the renderer, and never dom/ (which
> > resembles the interface only, and must stay as lightweight as possible).
>
> and because closing the renderer interferes it. Ok, fair enough.
>
> One other thing you mentioned was that the parser doesn't use the DOM
> interface. I guess you mean the DOM api, like appendChild and such. But
> grep'ing for these, do show up  alot of them, 

Mostly error handling. It shouldn't harm there.

> even in 
> ml/xml_tokenizer.cpp. Albeit for XML, looking at where it's used, does
> make one feel unconfortable adding the code we're talking about in
> xml/*. Eg. in XMLTokenizer::finish() there are all kinds of calls to
> closeRenderer(), and the remark is:
>   // Close the renderers so that they update their display correctly
>   // ### this should not be necessary, but requires changes in the
> rendering code...

Your concerns are valid. It looks like a little change, but may be way more 
invasive than it seems.
>
> Which means that if Paul moves his code, this (and more)  should be
> changed as well? If so, one could suggest this a post 3.4 thing, no?

That's why it's paramount to run the regression tests (this addresses you, 
Paul). I'm sure there'll be regressions. If they're too numerous, and/or too 
difficult to fix, we should indeed consider postponing the fix.

Alternatively, we could wait for Apple and check how they'll get rid of 
closeRenderer ;-)

mfg
 Leo

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20050211/585ce1f0/attachment.sig>