Several threads have started asking questions around this and we've been
rather quiet on our part, so I'll try to formulate some answers.
Up until the 4.4 release, Qt Jambi was a full time job for two
developers. This was for a number of reasons:
- The project was still pretty new and the bug load was steady, and
some of the bugs that came in were down-right tricky ;) Since the 4.4
release, we've noticed a significant decrease in bugs, so we seem to be
over the worst infancy. We've managed to fix most issues related to
library resolving, memory management, garbage collection and signal/slot
connectins by now, so the things that remain are simpler.
- Qt 4.4 added TONS of features. Each new feature needs to be ported
to Qt Jambi and even though this is ideally a matter of putting the
classname into an .xml file, there tends to be a bit more. With 4.4, we
had the problem that the generator didn't support namespaces, but phonon
required those. The Qt Concurrent API was heavily template based which
the generator didn't support so a lot of work was spend on the generator
to make it capable of supporting the features we needed for 4.4. 4.5
added very few features and I expect 4.6 to also be less extreeme, and
even if there comes a release with loads of features we're still in
better shape because the generator is more capable and the rest of the
issues are fewer.
- We do spend a significant amount of time in packaging, testing and
fixing the small things here and there prior to each of our releases.
Currently Eskil and I spend maybe one day a week on the project. Its far
from a full time job. Of course, we know all the details very well so
one coming from the outside and contributing his/hers first patch would
naturally spend some more time until they are up to speed.
When we're closing a Qt release, we tend to spend around from a week to
a month or two of time were we port all the new functionality to Jambi.
---
The codeline-illustration by Gregor is pretty correct. Qt Jambi is
primarily a C++ project and its all about understanding two individual
parts.
1. The Generator. Its the key to it all... It reads C++ header files
and uses the XML files to understand the API. In an ideal world, a
maintainer would never have to touch the generator C++ code, but just
add things to the XML files.
2. The runtime, the qtjambi directory. This is where the memory
managmenent is handled, GC'ing, GUI objects vs non-GUI objects being
deleted on main thread etc, but it goes hand in hand with the decisions
we made in the generator.
There is some Java code in the com.trolltec.qt package but its mostly
utility for stuff that got to nasty to write in JNI ;)
---
As for documentation, the innermosts parts of the generator and the
decicions we made during the porting, aka how do we solve multiple
inheritance is somewhat documented in the original whitepaper.
http://dist.trolltech.com/pdf/QtJambi_4.3.0_Whitepaper_A4.pdf
Eskil also wrote an article in the latest Qt Quarterly that I'll post a
link to once I figure out where it is ;)
---
As for help... We hope to get the hosting of the project up and running
and we'll probably do the patch-releases on the 4.5 series in the hosted
repository. We'll also be here for another year, answering questions, so
if there are people who are interested, they have a source for answers.
---
best regards,
Gunnar
___
Qt-jambi-interest mailing list
Qt-jambi-interest@trolltech.com
http://lists.trolltech.com/mailman/listinfo/qt-jambi-interest
