After discussions on the kde-core-devel mailing list, it is decided to lift partially the message, documentation and feature freeze for the KDE 3.5 branch (modules in branches/KDE/3.5).
The period of the lift will be until 10th July 2006 (00:00 UTC).
For the GUI strings (also known as messages), you can fix typos and make small changes to them. You can also add new error messages to improve error feedback to users. (Adding other kinds of messages are not allowed. In case of questions or doubts, please ask the firstname.lastname@example.org mailing list.)
For the documentation, the changes should also be rather small, except when fixing inaccurate or outdated documentation. You can also port documentation that has been prepared in the trunk/KDE modules. (If you change any documentation for the KDE 3.5 branch, please be sure that the change is also in the corresponding module of trunk/KDE. However you can do a little later, in case that you would not have much time during the lift period.)
In addition to that, the feature freeze as announced in  reapply to 3.5 branch during that period.
- Misspelling of the name is inevitable. In many languages (e.g. French) it's downright an error and unacceptable to start a sentence without a capital letter. So in official media it is always going to be spelt "AmaroK" at the start of a sentence, and possibly also in mid sentence. We're provoking this misspelling.
Here are some excerpts from the Helix engine desktop file:
- Comment[et]=AmaroKi plugin
- Comment[fr]=Module pour AmaroK
- It was a funny idea when Amarok was young, but nowadays it just seems quirky.
- The name is unusual enough not to require additional attention from special capitalization.
- Amarok is intended to be a software for all desktops, not just KDE. The capital K suggests that it's a KDE only application.
- Changing the name earlier is easier than later, assuming that our popularity is still growing.
- The Amarok logo does in no way reflect the current spelling. If anything, it looks like "amarOk".
- The various misspellings hurt name branding. There should be one brand "Amarok", and misspelling weakens our brand.
Next-generation desktop layer.
Seamless integration of hardware and networked devices.
Backend-independent multimedia layer.
KDE4 visuals and icon theme.
Extensible cross-desktop storage service for PIM data.
A service architecture to make chat and phone communication universally available across applications.
John and I are like starved dogs ready to take on UI complexity. We're bouncing ideas around, and will definitely be pushing some big improvements forward this summer. You don't have to worry about that.
So, for now the app remains KSysGuard. However, we can expect some interesting changes with this application (naming and otherwise) in the lead up to KDE 4.
|Commits||2156 by 193 developers, 5432 lines modified, 1759 new files|
|Bugs Opened||307 in the last 7 days|
|Bugs Closed||247 in the last 7 days|
Internationalization (i18n) Status
|British English (en_GB)||
Bug Killers and Buzz
|Allan Sandfeld Jensen||
|Aaron J. Seigo||
There are 84 selections this week
Need to distinguish the long/short forms of month and day names with
i18n() comments. In particular, this is necessary for the month of May,
which has the same long/short form. For consistency, I added i18n()
comment strings to all short-form names.
This is a backport from trunk, it closes bug #127011
Fix bug causing large lag in Full updates. .The problem was that I was recreating an Earth object inside the loop
over all copmets and asteroids, even though the Component itself
contains the Earth object. Fixing this caused the timing for
SkyMapComposite::updatePlanets() to go from 1.4 seconds to 0.03
seconds!.I also fixed some other minor issues in skycomponents.Note: I have committed the TIMING code that I had earlier sent to the
list as patches, so we can continue to use this information for
Trying to solve the remaining media support problems
- apply Coolo's patch for resolving media:/ URLs (from #93569)
- close CameraUI cleanly
- ask user if he wants to cancel the current operation when closing
- cancel operation when closing
- delete object when closed
- dont add new tasks to the queue when closing
- Allow only once CameraUI per CameraType object
- store current UI with a QGuardedPtr
- when a non-closed UI is found up and running, raise it, dont create a new one
- when camera interface is called via DCOP (probably from media menu)
- call KWin::deIconifyWindow
- call KWin::activateWindow()
- do not compute image dimension when listing from a UMSCamera. This is a _huge_
speedup, say 5 minutes to 5 seconds.
- do only compute image dimension if requested in UMSCamera
- compute image dimension if not yet done in CameraItemPropertiesTab
- various speed and crash fixes in CameraUI/CameraIconView
- do not crash when the first item was deleted from camera and then removed from IconView
- do not request exif for items which will be deleted
- IconView: allow to delay updating frequency a bit
- use a ched itemRect for CameraIconView. Its faster, but more importantly,
there was a crash with a pretty long story
CCBUGS: 127846, 93569
rewrote the networking code to use KNetwork:
* fixed #110915: KNetwork has better error reporting than the previous
code, now it says "connection actively refused" instead of "No such
file or directory"
* fixed #110838: this one was caused by an uninitialize variable, so it
BUG: 110838, 110915
Fix the address validator. 255 is obviously a valid part of an IP
One should also note that the netmask validator is incomplete. It
should check for a sequence of 1's followed by a sequence of
0's. Anything else is not a valid netmask.
Maximum idle time was again limited (my lunch break was definitly greater than 20 minutes). This time it seems the 'standby' state of DPMS was the maximum. Fixed this by expanding the workaround to also ignore the standby and suspend interval second.
Why o why does X emit activity when blanking the screen....
Last commit (hopefully) before INDI v0.4 is out. Will need to run a few
Please test this updated version to see if it works with your device(s).
I remove Astrophysics mount because there are no testers and it's been
broken for a while.
I noticed that KStars is crawling, I was barely able to do anything,
anyone else experiencing the same?
A new backend for oKular: DjVu:
* uses the DjVuLibre for loading the DjVu files
* can rotate the pages
* has a cache (with few elements) with the most recently used generated pixmaps,
to reduce a bit the number of pixmap generation
There could be some issues, but mainly it works nicely.
The holiday library now also supports
-) Multiple holidays on the same date (e.g. start of summer time and easter
on the same day, etc.)
-) conditional shifting of holidays (e.g. if a holiday occurs on a saturday
or sunday, it is actually celebrated on the following monday). Example:
small "May holiday" weekend on 1.5. shift to monday if saturday || sunday
The holiday data files still need to be adapted wherever such a system
applies (UK, Wales, Ireland are known from bug reports, but I don't
know which holidays are really affected).
It's incredible, how simple it is to hack these lex/yacc parsers (which had the fame that nobody in kde would understand them), once you take the ten minutes to print out a yacc tutorial and the 45 minutes to look at the code (and then the ~2 hours to fix the inadvertant bugs in the C code that come through the ancient pointer-increasing in C to walk through a list, when you are used to Qt-style list iterators ;-) )
Introduce a (not-pretty-but-working) toolmanager;
the toolmanager loads tools from all libraries that ship a .desktop file
with the serviceType KOffice/Tool.
They will then automatically appear in the toolBox of the application
and change the tool on the canvas should the user click it.
Replace playlist view tabs with Amarok style tabs. Advantage is that you can now hide/show the context view by clicking tabs twice, and it looks nicer (and possibly uses a bit less space).
Note: This is work in progress, the splitter doesn't work right atm, so the context view is semi broken. We'll try to fix soon.
rename SSI to OContact (OContact is to avoid a class with Kopete::Contact)
rename SSIManager to ContactManager.
ContactManager will become the class that holds all the details of all the
contacts that we know about within OSCAR. Eventually things will be switched
so that OscarContact and derivatives are derived from OContact and will use
methods from OContact for the backend support stuff, rather than fudging it.
we can now recognise incoming rendezvous messages, react to the cancel one, create new tasks for incoming requests (which immediately kill themselves), listen for direct connections, and die if someone actually does connect.
gaim chokes on acceping our request, though (claims we cancelled it).
also, this really needs a timeout to prevent zombie-filetransfers.
big commit. made undo work for command modifications :)
- workflowdocument private class has own header now
- new modifycommandaction class
- extended Command API to support undo/redo
- extended ValueList class a little
- internal classes take datatype names instead of pointers
Ok, this commit will change rsiwidget in a few ways. When there is no image found in the folder, it will gray out as in the kde logout sequence. Next to that, some artifacts seem to have been solved by these changes....
- place the counter in a box, maybe use the rsipopup for it
- make this behaviour configurable, next to the images.
- make this behaviour the default setting.
Please check for regressions and other issues.
I can not support transparant images as part of your image collection, but the aim is the title of the bugreport ;-)
Disabling antialiasing.I had enabled antialiasing for "non-slewing" draw events. i.e., when
the sky was in motion, the drawing was not antialiased. This explains
why non-slew draws took so much longer than slew draws.On my system, this change causes a non-slew draw of half the sky to take
0.3 seconds instead of 1.6. .It's too bad...I was really looking forward to the antialiased drawing
in Qt4. Hopefully, Trolltech will figure out a way to make it faster.
Perhaps the upcoming QGraphicsView class will help in this regard. For
now, I'll leave the antialiasing line commented out.You'll notice that the star colors look much more saturated after this
change; this is because the colors get washed out by the antialiasing,
so I had turned up the saturation to compensate..Hopefully, this commit and the previous one will make KStars usable.
digikam from trunk: use new dcraw thumbnail extraction method to render album items content and camera items content about all RAW files. This way will be always the more performant instead to use dcraw::parse method from digiKam core. Note that DCR raw files from Kodak camera works fine in this case.
IMPORTANT : you need to use last dcraw 8.21, else thumbs render can take a while !!!
Port of commit 548898:
Move the image transformations needed by the accessibility and
rotated viewmodes into the rendering thread. So now the GUI is
in this viewmodes as responsive as in the normal mode.
It also simplifies the code quite a bit, and and also uses only half
the memory because only one version of the page image is held
And as an additional sideeffect, the rotated viewmodes now workagain.
Since smootscale is time consuming, limit the loading of images to a bare minimum. Before this commit a new image was loaded when the config was entered (minimize() to make sure there was no active break) and when leaving the dialog another image was loaded (minimize() to reset timers). After this commit, only an image is loaded when needed. Hence the dialogs appear faster and when dialog is closed, rsibreak is immediatly picking up its task.
As the Summer of Code project should be "developed in public",
I can't wait with committing until everything works.
Therefore, import the beginnings of the C# parser into SVN.
The parser itself is currently just a skeleton, but the lexer is pretty
complete already. It only misses out on two things at the moment:
- Dealing with those ominous "non-keyword identifiers
that have a special meaning in the grammar"
I'll come back to this when I get to tackle them in the grammar.
- Filtering out pre-processor directives.
It's easier than in C++, because there are no text replacements,
only selective "turning on and off" for #if'd sections.
Not completely trivial though, but I already have a good idea
for how to cope with it, just need to implement it.
That's the upcoming task now.
Begin porting the KolourPaint paint engine to kpPainter, which abstracts all
_document_ drawing operations KolourPaint needs to do. And code simplifications.
Functionally nothing has changed except:
* Rectangles / rounded rectangles and ellipses are no longer 1 pixel higher and
wider than they should be and are now bounded by the dragged out rectangle
(rather than having the centres of their corners on the corners of the
dragged out rectangle)* Resizing documents now works again
* The fill style widget is rendered correctly again
This will eventually allow us to drop Arthur/Qt4 and use a proper paint engine. About kpPainter:
1. It draws separately on the RGB layer and mask layers (needed because without
XRENDER, drawing on the RGB layer does not touch the mask layer - this
is similar to Qt3 with or without XRENDER). This drawing pattern is
now centralised in ::Draw() instead of being spread all over the code.
To understand the pointer to member function syntax, read http://clarencedang.blogspot.com/2006/06/c-class-member-function-pointers.html .
2. Hides the mask layer before drawing on the RGB layer (with XRENDER, if it
sees a mask, while it will draw on it, it will also attempt to convert the mask to an alpha channel but
experiments show that it corrupts it; in any case, we don't want to
exploit XRENDER's alpha channel capabilities since a) you need XRENDER
b) accessing the mask when you have an alpha channel is a _very_ expensive
operation [http://clarencedang.blogspot.com/2006/04/so-whats-up-with-kolourpaint-in-kde4.html])3. As not all of KolourPaint has yet been ported to this new scheme
without XRENDER composition modes / alpha channels, transparency
still performs badly and is extremely buggy (see 2. above).
4. kpColor is supposed to abstract colour models for the document
5. kpImage is supposed to abstract image data for the document
(lossless, not tied to screen) but is currently a typedef for QPixmap
so nothing has changed yet - however, it is trivially easy to port from QPixmap
Ported Now Listening to D-BUS.
Of course it doesn't work because the media player aren't fully ported to D-BUS.
Will recheck when major media player will be ported to D-BUS and their interface be public.
Also whole Kopete compile with this commit.