[PyKDE] New PyKDE snapshot - help needed
jbublitz at nwinternet.com
Thu Jan 19 07:01:01 GMT 2006
I've just sent a new PyKDE snapshot (PyKDE-snapshot20060118) to Phil and it
should be available at http://riverbankcomputing.co.uk/snapshots/PyKDE as
soon as he has time to post it there (which is usually fairly fast, depending
on timezone differences).
The 2 major changes are that I've updated KConfigSkeleton (detail below) to
handle scalars the way people have requested (more or less) and I've added
the kabc and kresources modules, which are kdelibs support for KAddressbook
Both of the new additions were in response to requests from people on the
list. In turn, I'd like to request some help with those additions. The
KConfigSkeleton stuff needs both docs and a usage example program, the
KAddressbook stuff needs at least a good example of usage. I'd really
appreciate it if someone(s) could take some time to put some of that stuff
together (and also checkout the additions to make sure they work correctly -
I haven't actually tested any of this beyond compiling).
I also need feedback on whether this snapshot builds on various systems - I
currently am only setup to test on SuSE so none of this has been tested on
Fedora, Mandriva, Ubuntu or Gentoo. I don't expect major problems.
The next release will be PyKDE 3.15.0 or 3.15.1 - I've decided to sync the
PyKDE versions with the most recent KDE version supported, but I don't want
to go backwards to 3.5, so the 3.15.0 will have to stand for 3.5.0 until KDE
4.0 support is available, which will be PyKDE 4.0.
This snapshot brings PyKDE up to full KDE 3.5.0 support - it supports all KDE
versions from KDE 3.0.0 through 3.5.0. I've only backtested as far as 3.2.x.
It should also support Python >= 2.3.0, 4.0.x <= sip <= a fairly recent
snapshot (1/14?), and has been tested with gcc 3.3.1, 3.3.3 and 4.0.1. As
reported before, I can't get file concatenation to work with either PyQt or
PyKDE on gcc 4.0.1 - configure.py for PyKDE will automatically switch to
non-concatenated files for that gcc version (otherwise it concatenates all
the individual files for each module, which saves about 80% of compile time).
Since KDE 3.5.1 is just about ready to release, the actual PyKDE release
coming up may be 3.15.1 to include that. Either way, the offical released
version will be sometime in the next 2 weeks.
There were problems with the way KConfigSkeleton handled scalar types in the
addItem* methods and KConfigSkeleton::Item* class constructors. Each now
follows the pattern shown here using bool type as an example. For each
addItem* and Item* there could be two methods:
PyKDE uses : ItemBool* addItemBool (const QString&,
bool = 0,
const QString& = QString ::null );
Patch used: ItemBool* addItemBool (const QString&,
bool = 0,
const QString& = QString ::null );
PyKDE uses: ItemBool (const QString&, const QString&, bool, bool = 1)
Patch used: ItemBool (const QString&, const QString&, bool = 1)
In the addItemBool case, sip flagged the ambiguity between the signatures. In
the ItemBool ctor case, sip generated code for both versions, but the second
(original patch) version would be unreachable.
Since part of the argument for including these is more "perfect" KDE
compatibility, I've chosen the first (4 argument "official KDE" signature)
version in each case. Because of the ambiguity, existing code using the patch
won't break in the sense that it won't run - it just works differently. In
both cases, existing code written like:
x = addItem ("Name", False)
x - ItemBool ("Group", "Name", False)
will set the item's value but *will no longer set the item's default value*.
kabc and kresource
Both of these modules are essentially complete. kabc is missing the global
functions in ldifconverter.h, kresource is missing the Manager class. Both
will probably get picked up in a future release. I think the problem with the
ldifconverter stuff is just missing header info. The Manager class is a
fairly involved template class and I didn't want to hold up the release
working this out. PyKDE bindings for Manager will not be general in the same
sense a C++ template class is - they'll be tied to a concrete use, like kabc
- other uses of Manager will require additional code in PyKDE.
kjs support was dropped some time ago, and kdesu support a few snapshots back.
I don't plan on making either of those part of the official PyKDE. I haven't
heard any complaints.
I've done some code cleanup and added support for a few methods probably
nobody uses (mostly methods involving QPair as an argument or return value).
Tomorrow I'll be driving through Cascade mountain passes and the Columbia
River Gorge to Portland and back (12-14 hours altogether, longer if a lot of
snow in the passes), and Friday I'll be putting the hydraulic pump for the
end loader back on my tractor (I hope) - too greasy (and aching, most likely)
to do much computing. Somewhere in there I'll probably have to make some
I will, however, get to any bug reports and related stuff as soon after that
More information about the PyQt