[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: qt 3
From:       Marc Mutz <Marc.Mutz () uni-bielefeld ! de>
Date:       2001-08-23 14:01:06
[Download RAW message or body]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thursday 23 August 2001 10:57, Dirk Mueller wrote:
> On Mit, 22 Aug 2001, Charles Samuels wrote:
> > I see qt-copy has had something done to it, David is removing Qt3
> > stuff from koffice.  What's the status, then, on our switching to
> > Qt3 ?
>
> We will switch to Qt 3 as main Qt version for CVS HEAD when Qt
> 3.0.0-beta4 is released. Meanwhile Qt 3 porting is appreciated, thats
> why I imported a snapshot in qt-copy.
>

I have done kdenetwork/libkdenetwork (in CVS) and partly knode (not in 
CVS yet). I've run into problems with kdelibs headers, as expected, so 
this might not have been the last word for even libkdenetwork.

Something that regularly went wrong was kcharsets.h, and for obvious 
reasons: It concerns itself with QFont::CharSet, which is no longer 
there due to it being encapsulated inside QFont. What should be done 
with it? I don't want to work on it, but I need to know as early as 
possible because KNode, KMail and KMime do much handling with charsets.

Are there any plans for this? I've not seen someone posting about these 
kind of issues here yet. There should be an overall plan for things 
that need more fixing than a
s/QList/QPtrList/g 
and
#if QT_VERSION < 290
#include <qlist.h>
#define QPtrList QList
#else
#include <qptrlist.h>
#endif

Also: can the people working on kdelibs please start with the header 
files? This way, the people working on the other modules can start 
earlier to test at least compilation, and linking can come later.

> In my opinion CVS HEAD should till the switch still work with Qt 2,
> but if the majority wants to break Qt2-compatibility right now its
> fine with me. Its just a matter of a few days anyway.
>

I speak here as one of those that already have tried porting files to 
Qt3:
The problem when not using #defines is that you can go on 
incrementally. You have to convert a whole app in one go (in terms of 
CVS commits) and that makes it hard for multiple people working on the 
port f a single app. Also, you can't really test other changes when 
kdelibs isn't yet ported.

> The advantages of having both somehow working ( with #ifdef's where
> necessary) is that you can review the porting changes and see what
> got broken by it. if somebody just commits a patch to make it compile
> with Qt3 its a bit more difficult to follow.

This is true, too.

> Please note that the
> focus is NOT rewriting code or adding lots of features but to make a
> clean and straightforward (and correct port). KDE 2.2 isn't really
> stable as we know now, so KDE3 snapshots should become a good
> replacement in the near future.

The problem is here that there are a few major changes in Qt3 that 
possibly need much work to get right in the port. QRegExp's changed 
syntax together with it being implicitely used in QString::find is one 
example, another one being the changes to QFont.

I've yet to see a discussion on how to resolve such things when 
porting. If there are 10 people porting kdelibs, there will likely be 
up to 10 different appraoches.

You should say a word on this, Dirk. If we encourage people to do the 
port without discussing the big problems before, then it will end in a 
big mess. But then, perhaps this is the start of this discussion, so:

"What do we do with KCharsets?"

"Will there be a port-o-meter for each (sub)module, like in the 1->2 
transistion?"

I have defined three level of porting (well, four, counting "compiles 
and links" as fourth) each individual file. You can see it in 
kdenetwork/libkdenetwork/QT3.status. It would be cool if other would 
use this format, too.

Marc

- -- 
We have once again come full circle on the same basic question of
privacy on the Internet. If you have privacy, so does the person
sending around terrorist documents. And of course, we wouldn't want
that now, would we? [...] But what if governments, concerned about
mounting public pressure, decided to label protesters at the next WTO
roundtable, World Bank meeting, or G-8 summit as terrorists?
                  -- John Horvath: The Internet: A Terrorist Network?
                     Telepolis 2001/08/22 (#9350)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE7hQyj3oWD+L2/6DgRAoPZAJ90xVrEmBYJJ+Itl/xrzWokNHK+5wCfULTr
+8EpnbBXU6LVLpYrBF6y9Dk=
=TKqT
-----END PGP SIGNATURE-----

[prev in list] [next in list] [prev in thread] [next in thread]