Issue 90

Fixing Bug 94895:
The KDE4 codebase is much more efficient in the way it determines whether objects are onscreen, so we can now update the sky more often without stressing the CPU too much.

fixes for the terminal panel:
* assure that the correct directory is entered when reopening the terminal
* let the terminal be focused also when reopening it

This makes sure that if you create a new tab it will always inherit the current dir, unless the profile you use (either the default using 'new tab' or using a profile specifically) has a directory set. In which case it will go to that dir.

Fix the long standing limitation where multiple views on the same terminal session had to be the same size - the size of the smallest view. This means that it is now possible, for example, to have a large view on a session for browsing and examining output and a smaller view on the same session for input.

The underlying terminal does not support the concept of multiple windows, so the size reported to terminal applications is still that of the smallest view. Full-screen applications such as vim and emacs will therefore be the size of the smallest view on all views.

* Add a method in ScreenWindow to set the size of the window (in lines, the number of columns is still ifxed).
* Call this method in TerminalDisplay to set the window size when the widget is resized or when the screen window of a display is initially set.
* Fix ScreenWindow::getImage() and ScreenWindow::getLineProperties() so that they do not attempt to retrieve information about lines beyond the end of the screen.
* Fix ScreenWindow::getLineProperties() to always return result vector of size windowColumns(), though the number of elements copied from the screen may be less.

Fix KCursor autohide feature for scrollareas like KTextEdit:
the widget gets key events, the viewport gets mouse events, so we need an event filter on each one.

retrieve the LineEdit's frameWidth from the style rather than from a hardcoded random value.

fix linedits appearing too thin in oxygen style

allow the languages that don't have systemsettings translations to set the language too.

we need to mantain our own language list instead of using KLocale's because KLocale does not add a language if there is no translation for the current application so it would not be posible to set a language which has no systemsettings/kcontrol module translation

Patch idea suggested by Chusslove tested and approved by him too

- Properly update line numbers in the backtrace data structure all the time, not just when we're entering a context (so if one stops in the middle of something, the lines are actually correct)

- Make the callstack display actually usable, and let is jump around the code.
It doesn't change the context for the variable display. Maybe it should. Maybe it shouldn't. Not sure. But it sounds too much like a feature, so at this point I'll take "no" as the answer.

Set the panel's minimum size to its initial size so that removing applets doesn't decrease its size. This should be set to the size that the user wants once user resizing is implemented.

- Fix signal/slot connection for on-demand spell checking.
- Forward toggling of continuous spellchecking mode to the editor widget, which unfortunately doesn't work when it is in RichText format input mode.

- Woot! Fixing major breakage for funky usb devices. pulled back some removed code from KDE 3.5.6's kpilot for usb workaround code and fit it into the new threaded model. This fixes syncing for me on my Palm Treo 700p.
- Reformatting and cleaning up kpilotdevicelink.cc.
- New Private.h to hold the bits of kpilotdevicelink.cc that should not be exposed via kpilotdevicelink.h. Also, DeviceCommThread extends QObject and QThread, so we need to have automoc run on it.
- Switching from using QCustomEvent to using QEvent for device tickle thread to match what is done in kpilotdevicelink.

- Bringing changes over from kde3 branch.
- This fixes the Treo 700p (and other devices) by bringing back the pre-device-comm-thread approach, while still maintaining the fixes I brought in with the non-GUI thread for device comm.

Namely, we still use a thread to do all the low-leve comm work, but instead of looping, we allow 2 paths: 1) a timer fires, which means we couldn't re-establish comm, or 2) a QSocketNotifier tells us that there's data available on the HotSync port.

Fix some D-Bus calls to the groupware interface. Untested.
Once the groupware D-Bus stuff has been properly ported, we can use an auto-generated interface for this, so these kind of errors don't happen again.

Use qint64 instead of size_t for the folder size calculation, otherwise folders >2GB won't show up correctly on 32 Bit systems.

I hope I found all places were the conversion was necessary.
I couldn't really test this, as i don't have a 64 Bit system, but I should work.

Fix loading of .kra images (the imagename was set in the info block and then reset when an empty info block was loaded later on)

Fix from jakkapun (via Pattara Kiatisevi) for:
fixed the input method issue, added text normalization support

This makes Thai input work properly.

fix crap-tastic wiki applet. it's so bad i don't even care anymore, as we need Qt4.4/Webkit for it to actually do anything useful.

Finished settings dialog revamp :
- DHT and encryption now are on the BitTorrent settings page
- Stats plugin settings have been splitted into two pages

This results in a smaller config dialog, and so 152492 is fixed

removed KDE3-Context Menu as it doesn't work any more and crashes
added one icon to the chatwindow to at least be able to start OTR sessions

Reanimated the KCM Module:
fingerprintstable and own fingerprint combo box work again

Fix dropping a nested book into the root of the bookshelf.
No longer destroying users data. :)

Connect complete. Working on actions and minimal Kross widget interface.

Moved signals for starting bug query out of server list model to main window - query is started when query item gets selected.

Actually I guess backend API takes some shape, so I can start documenting classes and putting up some kind of DESIGN document with description of how they're connected

Implemented support for saving to remote urls

Added "Switch Tabs on Hover" option that allows to disable the auto-switching and requires instead a click to switch the tabs.

It does the same like the Kickoff@KDE3 KickoffSwitchTabsOnHover-option, see <a href="http://www.kdedevelopers.org/node/2620">http://www.kdedevelopers.org/node/2620</a>;

Add Bill Egert's train clock

Added "Number of visible items" option. This allows users to define the number of items that are visible which is used to define the height of kickoff. Default is 10 atm (probably set it lower to e.g. 8?).

A Now Playing engine, initially only giving info for Juk

We now have a rather crude nowplaying applet for Juk.

I invite anyone who wants to pretty it up.

cleanup, beginning of implementation of dvd, bufferstatus implemented...

Audio effects now work (and swapping them while playing)

Allow moving of even non-movable windows between virtual desktops using the middle mouse button (and also right button, newly, I guess that one is easier to use).

Now icons can be resized

Now with:
* default do-nothing implementations in Player (was PlayerInfo)
* volume support
* simple command (play/pause/stop/setVolume) support

- Put in the very basics of exception handling in the debugger, based mostly on the old debugger's code. Still very rough

Now the konsole part is autodestructed when no tabs do exist. This makes the desired effect on dolphin when typing "exit" on the terminal, the dock will be hidden.

New tag cloud. I put a lot of work into this one and now I am not sure anymore if a simple QLabel + some HTML would not have been enough...

Support for XEmbed-based plugins and Glib2-based eventloop.
Should make the most recent Flash work, if you get sufficiently lucky and don't run into any of the load of its bugs.

Latest kdelibs/kdeui/qxembed.* needed as well.

Simplify the pausing of CPU guard when blocked in debugger.
The pause/restore API in KJS::Interpreter is too subtle to use sanely in this context. Instead, merely ignore timeouts when blocked, and restart timers when unblocked.

This should eventually be replaced by a proper modality manager, so thinks like alert() work right, too, but I need to think it over some more. It is quite crucial, though, since there are quite some event looprecursion issues ATM.

kwallet support for the twitter plasmoid.

Allow rendering of a specific svg element id. You can still render them all if you don't specify a element name but now you can also use a specific element. Doesn't handle resizing.

Make battery applet work with multiple batteries. Patch by Maor Vanmak.

A step forward in making this clock sane and defaultable

Some D-Bus work in KMail:

- Remove the existing org.kde.kmail.kmail/kmailpart/mailcomposer.xml files, and let them be generated at build time instead. This is done by extracting the slots and signals marked with Q_SCRIPTABLE.
I had to adjust the CMake files of all users of the XML files to use the one from the build directory now.

Currently, the XML files are also installed. I am not sure if this is a good idea, as that implies we have to keep some sort of compatibility.

- Fix incorrect D-Bus path and method call in the summary view. Now the generated interface is used instead. Checking the accounts with the sync actions works again now.

- Some DCOP->D-Bus renamings in comments

- Some style fixes around D-Bus related lines

The D-Bus test of KMail and the Kontact KMail summary still work.
This needs the new FindQt4.cmake from kdelibs.
This probably needs a clean build.

I did not touch the groupware interface or the SMIME security config page (which talks to Kleopatra). This is left for KDAB to fix, but it would be nice to contact me before.

Interface changes in the legend tool:
* removed third tab ("Labels") which only contained the font entry. Unified title and label font properties to go along with the ODF standard and to simplify things.
* Changed tab order of legend type tool from data-type to type-data. After adding a diagram the user usually wants to change the type first.
* support for showing/hiding legend frame
* Unified/simplified legend tool widget layout

* Added a tool to select and modify datasets of the diagram. Now that selecting is implemented, dataset properties editing is to follow.
Note: For diagram types other than bar diagrams, this is still somewhat buggy. Only works for cartesian diagrams atm.
* Added ChartShape::repaint() along the way that doesn't update the entire KDChart::Chart. So it should theoretically be faster than ::update()

Support for changing the dataset color with the dataset tool.

Start of making the anti-aliasing of the gradient painter colorspace independent.
It's still buggy and slow, and parts of it should be going into the colorspaces (distance calculation and averaging), but at least this code compiles again.

lot of changes that can be summarize in "able to stich more than two images" (modulo a bug in the initialization of the optimization)

In more details :
- matching algorithms based on "groups match", faster and better results
- improvement in the harris detection, remove the threshold (or more like use a smaller value... like 0.0) on cornerness, this increase the number of interest point, and split the layer in zones to increase the distribution on the image
- introduction of the notion of models which are in charge of feeding the optimizer, and then output the transformation that can be feed to the blender - plus countless of changes to remove the limitation to two images

Added KPrCustomSlideShows which will provide presentations with any order of the slides, i meant... pages ;).

Started to merge all three existing tools into one, with an easy-to-use user interface and quicker access to more important settings.

Change the Calendar/day viewsChange the Calendar/day views&models to something more promissing.models to something more promising.

More work on the new unified Chart Tool:
* axes titles are now editable
* diagram grid can now be shown/hidden
* gap between bars and datasets can be changed
* legend can be shown/hidden

load ORA files

implement the selection of color in dynamic coloring program

Add a config screen to the Magnatune store service. Initially it just allows configuration of the ( not yet launched by magntune ) membership settings.

Basic scrobbler implementation in last.fm service, should basically work if the engine didn't have bugs that caused it to return wrong positions or that tracks aren't finished when they are ...

VideoWindow's context menu now works correctly, in both full screen mode and normal. The full screen toolbar doesn't work right, which isn't surprising given all the wacky stuff it does, its likely Qt3 dependent.

Add files for the new user interface:
-New button class (play,stop,mute...)
-Svg files in "skin/"

A little more work on the last.fm service, some radio support classes started but still incomplete.

completely re-work the layouting code.
this has more bugs (of course) but also more promise.

Add buy album option to Magnatune tracks. This time its actually connected from the start

Make 'buy' option in magnatune tracks and album menu become 'download' if you have a download membership. I cannot fully support free downloading yet untill I get some stuff added on the Magnatune side of things, so the download option stil takes you to the buy dialog

Changes :
- Added SOCKS support (v4 and v5, note: still need to do authentication for v5)
- Fixed some minor issues here and there

Added username password authentication for SOCKS v5

Added bonjour protocol for kopete.

Contact Management works. Sending/Receiving Messages doesn't

Added Facility To Compare Incoming Connections To Hostnames Of Contacts

FIXME: This Currently Shows only how many contacts are at given host.
We have to read the first few bytes of the stream to get the username
I'll have to start using libiris after this, I guess

Initial checkin of MS Cabinet file format archive reader

This class subclasses KArchive, and will (in the future :-) allow access to file that are archived in the "Cabinet" file format.

That format is documented, and there are other readers (e.g. cabextract) that handle it.

This still has a long way to go. I'm still struggling with the headers, and haven't yet done any of the decompression stuff.

Implement reading of non-compressed files.

I've temporarily borrowed a copy of KLimitedIODevice (from kdelibs/kio/kio), because it isn't public. This won't matter if KCabinet gets merged into kdelibs, but might require some harder thinking if not.

Add support for reserved data sections. At this stage we just throw them away, but if they were needed for anything, it wouldn't be hard to make it available.

Also add support for cases where the data is in more than one CFDATA block. We don't yet handle cases where a single file is spread over more than one block.

Add support for cabinet files containing compressed data.

This only works for MSZIP format compressed files (using the very handy KFilterDev decompressor, which implements the deflate algorithm).

This change uses code from the KZip file entry class.

Automatically add * wildcards to the start and the end of the filter string of the file selector widget.

This makes filtering easier and faster.

put the levenstein table on the stack, not the heap, and hence only allocate mem for it when needed, rather than having 65k*(number of files) allocated all the time.

(forward-port rev 444219 by je4d from 3.5 branch)

Profiling has shown that DocumentRange was one of the biggest remaining memory-hogs, especially on 64-bit systems KTextEditor::Range takes much more memory than the actual content because the data is split across multiple separately allocated blocks.

Also, it is inefficient to copy around, and DocumentRanges maintain a hierarchy that isn't needed at all, but also takes a lot of memory.

Solution:
Create efficient self-contained classes SimpleRange and SimpleCursor, that are used for each duchain object. Stop using DocumentRange in DocumentRangeObject. Instead, now each duchain object has one permanently stored SimpleRange, and a SmartRange pointer that is only nonzero when the document is loaded and has a smart-range.

The SimpleRange and the SmartRange are synchronized on access.

Result:
- No more thread-unsafe range access in DocumentRangeObject::textRange()
- Significantly less memory consumption
- More efficient serializability because the SimpleRange is embedded in the shared d-pointer.

By the way the duchain initializations were cleaned up a little, and some yet unnoticed bugs there fixed.

Committing optimization fix for dms::setFromString(), as submitted by Andrew Buck and improved by Luciano Montanaro.

Thanks for the submission, Andrew. I'm having trouble understanding how kstars could be spending so much time in this function. I don't see that it is called by any inner loops. I noticed that you patched against the 3.5.x version of the code (which is why you didn't find QString::stripWhiteSpace() in the Qt docs; that function is called trimmed() in Qt-4.x). But even in that version, I only see the following calls to setFromString():

KStarsData::processCustomDataLine() [used when initializing custom catalogs] KStarsData::executeScript() [used when parsing a KStars DCOP script]

dmsBox::createDms(), which in turn is called by FocusDialog and LocationDialog, in functions that should only be used when those dialogs are opened.

Do your gprof results indicate the calling sequence that led to execution of setFromString() ?

thanks again, and welcome to kstars development! :)

significantly compress update events: never repaint an item or relayout a task group more than 5 times per second.

note that we do an initial paint immediately no matter what to avoid appearing slow in the common case.

should protect us against insane applications.

* another obvious possible optimization ... i remember when gimp used to reset its window title on mouse movement resulting in a title update per pixel moved over in a document window ..... even though the title didn't actually change content. protect against such insane apps as well.

* if we don't have a group, call updateGeometry immediately (should be useful if we ever allow "bare" tasks, e.g. on the desktop)

Performance improvement. Not the last one. We still have much work here I think.

YEEEHA! Shared dpointer implementation for most of duchain (I'm going for the typesystem next).

Lesson learned: DUChain is _really_ picky about properly initialized member variables and right order of doing things.

Calculate include paths while still in main thread. This finally allows me to parse the entirety of kdevplatform and kdevelop, with all their included headers!

By using the main thread we avoid many race conditions, including with KConfig and KUrl.

I'm starting to think we may need a dedicated thread-safe api for kdevelop, or an auditing of which methods are thread safe...

Completely change the way forward-declarations work. Statically resolving them while building the du-chain simply is incompatible with re-using pre-compiled du-chains.

So make forward-declarations dynamically resolvable, using a given top-context as visibility hint.

In Declaration: Make the behavior of "internalContext" simpler, and add virtual convenience-functions "logicalInternalContext" and "logicalDeclaration" that resolve everything on the way.

Only browse through zip and tar files if they are local (otherwise a cryptic error message will occur).

We really must make this behavior configurable in KDE 4.1 - I'd suggest that per default no browsing through zip and tar is done (with an option turning this on again)

The breakpoint display doesn't seem to have any implementation at all,so just kill it for 4.0. At least saves some screen space.

Get closer to the original Kickoff terminology which says "Computer", and NOT "My Computer".

And no, whatever you might think: This is not a string change.

Blessed by Aaron.

Changed the "Recently Used" icon.
See before/after; <a href="/issues/2007-12-23/files/kickoffimage.png">http://kross.dipe.org/kickoffimage.png</a>;
Hope it's ok (aka matches more to the other 4 icons)?!

I think zoom would be much more useful if wheel could control it.

Remove dead code, including the non-functional scroll lock code. Scroll lock support will be re-implemented in the terminal display widget post KDE 4.0 if there is a demand for it.

don't make the background kcm and it doesn't actually affect the background anymore. kdm stlil relies on the code there, though, so it can't simply be removed yet.

the currnet background dialog in plasma will get pulled out and made into a new kcm.

Implement Applet::shouldConserveResources(), a method to help plasmoid developers to know whether they should conserve resources.

Peer reviewed by Chani and Richmoore.

Applied a slightly updated version of the patch discussed on kde-core-devel to make scrollbars apply to the module content only, not to the bottom button bar or the left icon view.

Incorporate Fredrik's delegate. Now the speedcrunch-ripoff lookactually looks as nice as the original

Try to place unpositioned applets in areas that aren't occupied using a very simple placement policy.

darken the highlight color for KDE manuals in the browser, as it was almost invisible on cheap displays (like mine ;-).

Got no objectionis from kde-i18n-doc and kde-usability against this change.

Show debug warning if a broken icon has been requested.

Discussed on kde-code-devel.

Note: ifndef NDEBUG does not work yet. It will make sense as soon as this define works again. If you compile in debug mode, warnings will be shown for every broken icon at every request. In release mode, warnings will only be shown once; after the 'unknown' icon will be in cache.

Finally, Kleopatra has an application icon. Thanks, Thomas Moenicke!

- Automatically lighten the KMail system tray icon if the unread count is displayed.
- Change a few colors to use KColorScheme

Patch by Jakob Petsovits, thanks!

copy the experimental open raster loader in a more generic framework to read/write and experiment with open raster

add initial experimental untested and unconnected to a mimetype filter for OpenRaster (long-term) Archive file format but with only saving implemented

Tag Amarok 1.4.8.

It's decided, rename Video Player to Dragon Player.
I've always been a Konqui fan, I had a Konqui shirt years before I started developing. :)

just for fun, try to bring back the animation in the tree views to see if it has become less buggy, let me know if I need to disable it again.

Remove GeoIP-things...(because of licensing :( )

- Add bg gradient to menus
- Finalize the look of selected menuitems

adding kdewin-menubuilder: This program is for internal use only. Integration into windows emerge.py and kdewin-installer will follow. See kdewin-menubuilder --help for help.

The menubuilder isn't ready yet, I am still searching for a solution about the Categories (maybe I can use them for the folder structure?).