Issue 272

always explicitly set the text of the first column

otherwise on data update, it just shows whatever was there previously, which
is wrong and obviously highly misleading

Extend test, verify the function type.

This shows that something is wrong with the JSON test suite...

Use proper target exe path calculation for CTests

Calculation of the executable path is done through the
ProjectExecutableTargetItem which respects RUNTIME_OUTPUT_DIRECTORY
target property and CMAKE_RUNTIME_OUTPUT_DIRECTORY variable values.

Addtionally Test::executable is not modified if kde4_add_unit_test
detection logick failed (target name calculated in the detection block
dies not exists in the code).

REVIEW: 108556

Fix AndroidRunner stop action.

When the stop button was pressed nothing happened

Change-Id: I58a05aa3a4ea990595298cde0a330b382b65b035

Always set type on function declarations.

The return value of the visitor functions only tells the
visitor whether to continue looking at the children nodes.

Fix SVG export

Some code was disabled, because it required
Qt 4.5, which we long assume.

Thanks Leonardo Barichello for spotting this!


FIXED-IN: 4.10

status is unreliable, check for nan explicitly.

Losing the GPS fix often results in gpsd reporting status as ok, but
reporting nan as position. These invalid coordinates result in broken
position data (nan as position, random painting on the map) and likely
also crashes.

Fixing mistake in fetching of DSS imagery for deep-sky objects.

Now, the DSS image size is really determined by the size of the
deep-sky object.

When querying DSS, use POSS2/UKSTU Blue plates. They are much better.

Ideally, we should give the user the choice.

resource editor: don't crash when deleting the first keyboard layout

The root problem of the issue is the kdelibs bug 303228
("KCategorizedView is calling visualRect recursively"), this commit
provides only a work arround for the issue by temporalily deactivating
the categorization of the resource list when deleting a resource.

Fix bug with NGC / IC multiplicity suffixes.

NGC / IC objects sometimes have multiplicity suffixes.
eg: NGC 4945A is the tiny galaxy near NGC 4945.

Because we did not read these values correctly, RA values for objects
with suffixes were read incorrectly and assumed to be zero. This bug
was reintroduced while moving to KSParser.

Also, the suffixes were not included in the names of the objects. Now
they are.

BUG

Merge branch 'load-img-with-wrong-extension' into KDE/4.10

REVIEW: 108470
FIXED-IN: 4.10.0

add info about Table Selection Tool and that browse tool mode is necessary for Inverse Search

Try to make it fail on jenkins too

Seems the path is not being utf8 and thus stuff does not fail
Don't think this will do anything, but won't hurt

Fix for bug 312823 - There is always a "undefined[Finished]"
notification when gwenview starts up:
Don't try to load an empty KUrl.

REVIEW: 108565

FIXED-IN: 4.10.0

Some cleanup in docs, fix EBN warnings

Rewind buffer in case a bad Qt image decoder moved it

Fix decoding jpeg files with a .gif extension.

Add a test (that needs pdflatex) that tests synctexpdf

The second one fails (patch is coming in a moment)
This is just to try what jenkins says of all this :D

Don't crash if a null KFileItem is passed to
ArchiveUtils::fileItemIsDirOrArchive

REVIEW: 108561

check if list is empty before to try to get first item. This must prevent internal Qt crash

Use pathOrUrl rather than prettyUrl, gets rid of file:/// when overwriting file

REVIEW: 108544
(cherry picked from commit 7687aba7026799c43cec4fa289bcd527e8ab8f6c)

Fix dirwatching being stopped when using KIO::moveAs().

CopyJob calls stopDirScan to avoid notification storm (it uses KDirNotify
i.e. DBus to notify of the operation instead), and forgot to call
restartDirScan in the precise case of direct renaming. Moving a '}' fixed it.

With many thanks to Frank Reininghaus for another great unittest!


FIXED-IN: 4.10

Make plasma widgets respect plasma theme colors

Patch adds ButtonBackgroundColor to PaletteHelper. Not sure it's the best way for fixing the bugs, but it works :-)
Problem is only exposed in combination of light colorschemes and dark plasma themes which don't have a light button (so have the same button text color as for window text).
PAtch by Hrvoje Senjan

REVIEW:108525

Ensure we destroy our test windows

KEditTagsDialog: Use the uri not the label as the unique identifier

The EditTags dialog used the Resource::genericLabel() as the unique
identifier when accessing tags. It would save this genericLabel in the
Qt::UserRole of the QListWidgetItem. This caused the following problems -

* When removing a tag, the tag would be identified via Tag(genericLabel),
this will NOT always give the correct tag back. Mainly cause
Tag(QString) operates on the nao:identifier not the genericLabel.
Quite often the genericLabel and the nao:identifier are the same, so
this wouldn't always be a problem.

By using Tag(genericLabel) we try to delete a tag with nao:identifier
genericLabel, which didn't actually exist.

* When saving tags, it would create each tag with the Tag(genericLabel),
this would normally be fine, unless there exists a tag whose
nao:identifier and genericLabel are not the same. In that case the tag
would be duplicated, and since the user only sees the genericLabel, it
would result in 2 identical tags being shown.

* When saving the tags, it would try and re-create each tag which was
being returned. This resulted in a lot of unnecessary Tag::setLabel
calls which cause unnecessary writes, and slow down the entire
process. This can get especially slow when marking a file with a large
number of tags. We now only set the label of the tags which do not
already exist.

a whole bunch of fixes, implemented WinOpticalDrive::supportedMedia()

Removing remaining problematic MenuIndex entries from profiles

don't add a type to the watcher

if we narrow down the types watched, when properties that are catched by and inverted term change, they won't be signaled if the object where those properties belong is of the wrong type:
example: isRelated property for activities, if we do a query asking for items of type FileDataObject

don't hide popup when under mouse

restores behavior of popups in 4.9

FIXED-IN:4.10

Rewrite of KWin's Screen Edge Handling

This rewrite is mostly motivated by the need to handle multi screen
setups correctly. That is have edges per screen and not for the combined
geometry. Also porting from XLib to XCB has been a motivation for the
rewrite.

The design of the new ScreenEdge handling is described in the
documentation of ScreenEdges in screenedge.h.

In addition the following changes have been performed:
* move configuration from Options to ScreenEdge
* add screen edge information to Workspace::supportInformation (obviously
replaces what had been read from Options)
* have Workspace hold a pointer to ScreenEdges instead of an object
* forward declaration of ScreenEdges in workspaces.h, this explains the
seemingly unrelated changes of just another include in some files


FIXED-IN: 4.11

Fix window not showing on correct desktop at startup


FIXED-IN: 0.1.2

Use KConfigXT in KWin::Options

For the time being the current design of Options is more or less kept to
not have to adjust KWin code all over the place. Also for some parts the
generated class from KConfigXT cannot be used due to inter-settings
dependencies defined in the setters.

Options now holds a pointer to a Settings object which is generated from
KConfigXT and uses it to read the default values and the individual
settings. This means the static default value methods are dropped and the
variables are initialized with a normal default value (all int 0, all
boolean false and so on) in the initializer list. Afterwards the values
are set to the correct default value through KConfigXT.

So far for the first step only Windows category is using KConfigXT.

REVIEW: 108572

Also set the font's style name when batch-changing all fonts.

This fixes a regression created by Qt 4.8: calling KFontRequester::setFont()
ends up calling the nearestExistingFont() helper function, which in turn
calls QFontDatabase::styleString(). In Qt 4.8, QFontDatabase::styleString()
calls QFont::styleName() before falling back to the old code path.

In our case, we ended up inheriting the style name of the current "General"
font.

In practice, this means that if the current "General" font had a style such
as "Regular", it was not possible to apply another one such as "Light" or
even "Bold" to all fonts because the other properties we set were
disregarded in favour of QFont::styleName().

Changing each font individually worked fine because it uses a different code
path.

Make the "A folder cannot be dropped into itself" message less intrusive

Showing this message in the KMessageWidget above the view, which means
that the view contents are moved down, can be extremely annoying
according to user feedback. Just showing the message in the status bar
is probably enough.


REVIEW: 108483
FIXED-IN: 4.10.0

discard mouse events outside the contents rect

in this way applets won't steal events from the handle buttons anymore

Fix a number of inconsistencies in state handling.

Make changes to the number of active threads at only one time, while
holding the lock.

Add basic documentation for StateImplementation.

This fixes doxygen´s class hierarchy view.

CleaningJob: Do not clear files whose url does not exist

RemovableMedia files are no longer stored as filex:/uuid/filePath. They
are stored as normal file urls. We do not want them to be removed when
running this cleaning job.

Fix "switch user" action not doing anything

Thanks to parcox for the patch.

@parcox: would have committed it with you as author, but KDE Git server
disallow authors whose name does not match "$firstname $lastname"


FIXED-IN: 0.1.2

don't stopMousePolling when not started

REVIEW: 108536

add unit test for crash 313769

enable dynamic word wrap

1. open kwrite, resize window to 900x1000 pixel
2. open test file "data/bug313769.cpp"
3. <cursor down><cursor down><ctrl+shift+end><ctrl+shift+cursor-up>
4. <ctrl+pos1>
5. <ctrl+shift+minus>
6. undo, redo, undo

Tags kioslave: When deleting a file, just remove the tag, do not delete the file

Cancel in-flight dbus queries to avoid leaks.

If close() is called on the query service client while dbus queries
are still in flight, they were left dangling. This change makes sure they are
properly cancelled. Without this, all such dangling queries are maintained by
nepomuk and updated constantly, causing extreme loads, over time.

REVIEW: 108550

VirtuosoInferenceModel: Cleanup

* Remove the type visibility code. We no longer rely on type visibility
for removing the ontology results.

* Temporarily disable the ontology graph group creation. It was created
in order to easily remove ontology results from Nepomuk queries.
However, virtuoso crashed a lot with the usage of graph group
exclusion.

Therefore graph groups aren't really used anywhere right now. Maybe
with future virtuoso versions. So, for now the code stays, but it is
not called.

Also added an appropriate warning in the function.

Tags kioslave: Allow recursive deletes

No point individually removing the tag from all the files it is tagged
with. Just removing the tag will do the same. Additionally, we do not
want to remove other tags when removing a tag.

Prefer a more meaningful binary name than a wrapper like "kdeinit4"

It is a little confusing for users, when they see "kdeinit4" is
mentioned in the "Detail" and backtrace header while it is "kdeinit4
krunner" that crashes.

This also improve consistency, since the title of DrKonqi dialog also
says it is krunner that crashes.

some screen locker visual fix

1. string in qml doesn't have "empty" property, use length to check it's
empty or not.
2. AccelButton's string doesn't have &, which will result in unpressable
accel key when process i18n string.
3. AccelButton will display & if showAccel is false, here we introduce a
new property for normal string
4. fallback to loginName if fullName is empty.
5. console.log() in AccelButton is removed.

REVIEW: 108559

don't shortcut Above restacks

just because we're above the sibling.
There could be another sibling to get below.


REVIEW: 108609
FIXED-IN: 4.11

Kinfocenter: use unique names for some subdevices

The names for disk devices are now the mount names (sda1) instead of
using the product name, which is same for all the partitions.

REVIEW: 108529


Signed-off-by: Tomáš Chvátal

Change tab and window icon when someone is typing or there is unread message

Tabs will have 'unread-message' icon when there is an unread message.
When there are no unread messages, but other side is typing, it shows
a 'typing' icon. Otherwise contact's presence icon is used.

The window will have 'unread-message' icon when there is an unread message
in any tab. Otherwise will show 'typing' if user in current tab is typing.
Otherwise it shows current tab presence icon.

REVIEW: 108500

FIXED-IN: 0.6

Apply patch from Torsten Grote "Don't add line breaks into URLs"

Thanks


REVIEW: 108254

Fix Bug 313652 - pimsettingexporter shows blank icon in its "About PIM

Setting Exporter" dialog

Use kontact icon

Fix Bug 313538 - KMail takes about 25 seconds to refresh background when

you click on "Serverside Subscription..." in IMAP account.
FIXED-IN: 4.10

No i18n in the global namespace

You need a kcomponent for i18n to work, and if it's in the global namespace it gets inited before anything can happen

Please do not ignore
"KGlobal::locale::Warning your global KLocale is being recreated with a valid main component instead of a fake component, this usually means you tried to call i18n related functions before your main component was created. You should not do that since it most likely will not work"
in the future

Add a dlopen hack for disabling VSync

On some systems QLibrary cannot find needed OpenGL functions, so
we need to use pure dlopen for this.

We should port this code to GLEW, when GLEW 1.7 will become a mandatory
dependence.

Added a workaround for OpenGL bended lines

There are two reasons why the lines might become bended:
1) swapBuffers() call waits for the display VSync for about 6 ms, all the
events seem to be dropped during this time.
2) When painting very *huge* circles with very small brush
(circles ~ 3000px in diameter, brush ~5px) some another factor
appears. I cannot tell for sure, but this is, probably, texture loading.

This patch fixes only the first (!) reason of line bending.
First, it asks Qt to use single buffered contexts. Then, if a single
buffered context is not available, it asks openGL to disable VSync.

Changed default shortcuts as suggested by David Revoy

This patch fixes most of the shortcuts mentioned in the bugreport except:

"Save As..." -- this is a standard KDE shortcut, so you can configure
with a Configure Desktop dialog

"Fullscreen" -- this is a standard shortcut as well, but you can't easily
set it to "Ctrl+F" due to bug 313621 in KDE shortcuts
system.

Alternatively keep an updater in stash itself

This will happen when we haven't got the imageshare UI. At the moment
that means when we have QML. Arguably we can now remove that from the
imageshare plugin and just use this instead. (might happen later)

workaround for the focus lost, since the slot is called when the editing is finished or we lost focus we need to check if we really lost focus before hiding the edit

Made the Freehand Tool support intermediate undo's

Now you can use Ctrl+Z during transformation again!

Fixed a flickering bug caused by disabling Double Buffering

Now disabling the Double Buffering is the last resort for the
workaround of bended lines. First, we try to disable VSync and only
if it fails we disable Double Buffering.

Make Compile on OS X. Apparently Xcode gets confused between Meta::Base and MemoryMeta::Base

i18n: Use argument for track count on splash screen.

Fixes a bug reported by Albert Astals Cid on behalf of the translation
team. Also has the effect on my system of inserting thousands separators
like normal, perhaps I should have made this switch earlier...

Albert, please let me know if the translators desire this to be
backported, as it stands now I have no intention on breaking i18n freeze
for 4.10.


FIXED-IN:4.11

Use consistent naming (backend), use the full form

Playlist sorting by type does not work


FIXED-IN: 2.8

Inline and simplify the branch rules in the common kdesdk file.

common-kdesdk-rules was including common-kde-branches-rules,
which has lots of unnecessary kdepim-specific stuff.
I inlined it into common-kdesdk-rules and did a lot of simplification.

I have verified this commit doesn't change the conversion results.

Fix moving & resizing of timeline zone sometimes broken

EngineController, Fadeouter: improve fadeout behaviour by using RAII principles

Note that this needs phonon-gstreamer bug 313551 solved to unleash the
full power.

BUGFIXES:
* Fix bumpy fade-out when first used.
* Fix subtle bugs when a lond fade-out is initiated near the song end.
* Don't try to fade-out with phonon backends that don't support it. (vlc)


FIXED-IN: 2.8

Fix a bunch of QML Warnings

Add workaround to prevent crash.

Initialize variable here to fix Metric edit box in VPN's routes window.


FIXED-IN: 0.9.0.7

Scroll when dragging contacts

As we override the drag+drop of the contact list the default scrolling in drag move event was not called.
This patch fixes that


Reviewed-by: Martin Klapetek

Fixes javascript error for word replacement from the spellcheck popup menu.
Restores Auto Spellcheck option in advanced settings

Copy the exit status of the child dpkg process to our return value, otherwise commit errors won't trigger an error message.

Use QLatin1String to avoid implicit conversions from const char*. (Fixes build with QT_NO_CAST_FROM_ASCII)
Also some minor const correctness things.

Properly fetch screenshots

It will be broken since we're using local naming vs. debian's, but it works
in some cases...

Provide a proper changelog for Bodega resources

Uses the new API that Aaron just added (thanks :) for extracting the
changelog data.

fixed link to installer log file in error dialogs (Ralf Habacker)

use windows eol in installer log file

Save touch/pen mode into profile after using the hotkey

The tablet will save the state for touch on/off and stylus/eraser
absolute/relative mode in the profile when the global hotkey that
changes this are used.

Fixed wrong assumption that only lower case package names are used:
Language packages may use upper case language postfixes eg. _BR.

rainbow colors for variables

Simplify Code: build types directly in the DeclarationBuilder.

While the context builder is decoupled from the decl builder to
be reused in the use bulider, it does not make sense to also
decouple the typebuilder from the decl builder.

Merging both simplifies the code and makes it faster as well due
to less required locking, indirection and and lookups. We should
do that for all language plugins, really...

Also use rainbow color for global variables.

Actually, we use it now for every non-function declaration. Once
we start to support properties we have to think about how to handle
that. the function-based prototype OO system will be interesting
to handle here.

Three improvements to the D-Bus method getObjectDataXML

1. Return a Constellation field which indicates which constellation
the object is currently in.

2. Move the names into text elements for uniformity and easier parsing

3. Fix the value of "Minor_Axis" when the object is circular.

Support for moons in the Celestial bodies selector added

REVIEW: 108392

First approach to matrix implementation

Only with + operator implemented, but with the parsing and type checking
in place, apparently...

Adding two new objects to the NGC / IC catalog:
1. Melotte 111 -- the Coma Star Cluster
2. Collinder 399 -- the Coathanger / Brocchi's Cluster

Also adding common names for many NGC / IC objects. Adding alternate
common names for some objects.

FEATURE

EclipsesPlugin: Enable lunar eclipses

OsmAnnotate: draw edges for areas

support calling the greeter with immediate switch user UI

IndexCleaner: Add another removal query

Add a query to remove old strigi data where a res was part of another
resource which had the uri of the form </home/something..>.

FileMetadataProvider: Index the file if it has indexingLevel = 1

With 4.10 the file indexer is only called when the user is idle. This
means that the user could select a file and its information will not be
shown in the sidebar unless it has been indexed. This is not ideal. The
users want to be able to see the indexed information. They don't want to
have to wait for the indexing to complete.

Now, whenever the FileMetadataWidget displays the results of a file
which has not been completely indexed, it calls the nepmukindexer
process to index it completely.

ffmpeg analyzer: Support matroska videos

add switchUser to ksmserver's dbus API

this will allow us to move switch user functionality from krunner to ksmserver
as ksmserver manages the sessions (ergo its name) and the lock screen, this is
a logical place to put it. it also means we end up with just one UI (in the lock
screen greater) that is fully themable (in the QML sense: interaction, looks, etc)
for user sesssion and guarantee that the current sesssion is locked before we
switch to another session

IndexScheduler: Inform of status change on suspend/resume

Take a different approach for "zooming icons", imho a zoom slider is not needed.
And the only place I could imagine the zoom slider to fit in is left of the Ok/Cancel buttons
which is not possible, you can only add PushButtons there.
So, I add a "Zoom icons" checkable button (will have to evaluate if/how I make it remembers
if it is checked) that makes the icons show in 128x128 or 256x256, so if you want to see
inbeforehand how the icon will look in huge sizes, you can check this button.
But a zoom slider with multiple sizes is not needed imho. It's quicker applying the icon
and then playing around with the icon size in Dolphin rather than adding a huge code path
into the dialog.

Added geometry and Tooltip Feature (code cleaning/formatting required)

Add fileIndexingDone dbus signal

In order to start the webminer right after the fileindexer did his job
this signal is needed.

Usually the service is started via the eventmonitor after the usual 2min
idletime. If there are no files with kext_indexinglevel 2 the scheduler
is stopped till the next 2min idle time kicks in and the scheduler is
executed again.

Now with this signal the webminer scheduler can look for new stuff once
new files are indexed.

This is extremly useful if indexing on Normal mode is enabled in the
nepomukstrigirc.

With this signal right after downloading a pdf/video the fileindexer
extracts its part and the webminer can be started right away, thus the
user can see immediately what nepomuk can do for them.

REVIEW: 108469

Add a way for source models to expose more actions

Add dbus interface for interaction with the service

The WebMiner service can now be watched/controlled via dbus.

Available is:
* suspend
* resume
* status information

Still missing is output for the percentage signal that would give more
information about the current status of the indexer.

Add support for KRunner actions

provide support for calling up the switch user UI

Add run associated application action to handle

Nicer handling of QML loading failures

systemd session management support in the workspace

CK code is preserved but systemd is checked first

Makes switching users possible

REVIEW: 107611

Implement overwriting of existing values in the Mac kwallet version.

REVIEW: 108578

add actions to wrap current word or selected text into specified open/close characters (like: (){}[]<>"). incrase indentation level in case of multiline selection

Activate monitoring on symlinked files

Previously kate would monitor the symlink for changes
and ignored changes made to the actual data. This could
cause data loss if changes were made to the file behind
the symlink while editing in kate.


REVIEW: 108555

Publish the ontologyUpdateFinished Signal on DBus

Notify when all ontologies are updated. This is sound with notifying
every ontology that is updated.

REVIEW: 108437

Create a clickable link for reporting crashes of drkonqi itself

That should be more user friendly than expecting users to open
bugs.kde.org manually and find the drkonqi product.

power: hide unavailable sleep modes

Many thanks to Philip for pointing me to the code to copy from!


FIXED-IN: 0.1.2

connect to the show user switching signal and respond to it

Fix kconfigcompiler_test by porting to QFINDTESTDATA

It previously did not find the required files since CMake set TESTSRCDIR
wrongly. It was surrounded by quotes and lacking a trailing /

This did not cause test failures since QCOMPARE/QVERIFY was not used
directly in the test function, so it simply printed messages instead.

REVIEW: 108560

ResourceData: Set the nao:prefLabel for new tags on store

When a new tag is created only its nao:identifier is used. However, most
people using Nepomuk seem to rely on Resource::label() to get the tag's
label. There is no convenient way to access the identifier.

Therefore, we might as well set both the label and identifier on new tag
creation.

Re-enable direct nepomuk search for completion again by default.

dfaure, who originally introduced the option and advocated it be off by
default now thinks it's fast enough, after the recent changes, to be on
by default at least in master, for experimentation. I agree, this makes
it so.

Use a faster contains expression for "start of word" matches.

For the special case of address completion, we are really only
interested in matching at the start of various parts of the name, first
name, email, etc. So for that case, use bif:contains('foo*') rather than
the much more expensive regexp match. Speeds up addressee completion
something fierce for a minor loss in functionality, namely arbitrary
substring matching.

drafts improvements

- better "need-to-save" detection
- autostore in cache, not temp
- fix file permissions
- delete sent drafts
- use QLatin1Spam
- more comments
- warn user that attachments are not saved

offline: allow opening mailboxes which are cached when offline

fixes #592

GUI: reduce the interval in which "today's messages" are recognized as today's

Finalize the Fill Config Widget behaviour.

Expose backgrounds in the resource poppup action.
Keep internally the resources.
Update the widget, and it's icon in the right way.

Add sketch installer files

use cross cursor and foreground color in calligraphy tool
, 313636
REVIEW:108562

Ported Transform Tool to strokes

There is one regression: the actions inside a single transform stroke
(that is until you applied the transformation) cannot be undone yet.

The undo of the transformation as a whole works perfectly.

,313294,310559,297929

feature: Buttons on dashboard widgets to discover the contextual menu

Introduce 3 new user-configurable opt-in columns
in the homepage's accounts table reflecting the
number of not marked, cleared and not
reconciled transactions per account

REVIEW: 108150

Implement Undo command for Paragraph Settings


REVIEW: 108491
(cherry picked from commit d4869442d2ec15da9130f81d06b9c4c41f8e791f)

Conflicts:
libs/kotext/KoTextEditor.h
libs/kotext/KoTextEditor_format.cpp

Added simple implementation of the texturing with real paper effect

Integrated new playlist module.

Activating write single file disables the playlist capabilities in the interface.

Handle all types of connections

Added option for opening new tabs next to current (default to true). Otherwise tabs created at end of list.

First draft of proxy model adding Telepathy presences

The idea is to provide presence and capabilities information directly
from Telepathy, which gives us instant changes and one less layer
for the data to go through and even easing the Nepomuk load. This
is done in a proxy model, so if we don't need the real-time presence
updates, we can use the PersonsModel directly.

Implement incomplete macro support

Use KTp::Actions for starting audio/video call

Improve icon management code

In the 1.x way, IconManager notified all the tabs about icons changed.
This way, just the interested tab is :)

Merge branch 'discover-mainwindowification'

This branch improves the integration of muon into the rest of KDE workspace
and applications.

make sure that we can list packages.

it's an important milestone, since we already get the list of packages from
PackageKit

Move setupActions to MuonMainWindow

This way every window can disable/enable all actions at the same
time without depending on QAptActions.
This lets us finally detach from libmuonapt, now muon-updater
should be able to work with any backend.

Bring back updates showing in the toolbar

Since the porting to KToolBar, we lost the feature where we'd see how much
packages we can update from the overlay icon.
Now it's shown together in the button text. It also helps since it says
that those are updates, so you know what's the button about.

Create a generic backend updater

Basic code completion support.

It just lists all declarations from the current context,
but now there's all the boilerplate code and an example of
how it works. Thus, it should be much easier to implement
better completion from here on.

Support global shortcuts to switch screen mapping + TabletArea

The screen selected screen the user wants to use can now be switched
via shortcuts between "All screens"/"Screen 1"/"Screen 2"

Also the used TabletArea calibration can be set individually between
the selected screen mapping and will change accordingly. If the screen
is toggled the stylus/eraser will be set to Absolte mode automatically.

For the implementation the ScreenMapping property is now depricated
Instead the ScreenSpace can have the values
* full
* map0
* map1
* x y w h (for part of the screen mapping)

In full/map0/map1 mode the actually screen size as reported by qt will be
used automatically without looking into the config.
Whenever the screen mapping is changed the values are saved in the config
to be remembered for the next reconnect.

Also the Area value will be changed as copied from the
* AreaMapFull
* AreaMapPart
* AreaMap0
* AreaMap1

properties that store the tablet calibration configuration for each
screen mapping.


DIGEST: This patch makes working with multiple monitors a lot better

Add benchmark comparing different possible containers for the TypeRepo

This was discussed on IRC and I thought it might be interesting. Here
are the numbers for a RelWithDebInfo build on my machine:

RESULT : BenchHashes::typeRepo():"vector":
0.00012 msecs per iteration (total: 67, iterations: 524288)
RESULT : BenchHashes::typeRepo():"vector-raw":
0.00012 msecs per iteration (total: 66, iterations: 524288)
RESULT : BenchHashes::typeRepo():"qhash":
0.0011 msecs per iteration (total: 75, iterations: 65536)
RESULT : BenchHashes::typeRepo():"qmap":
0.0019 msecs per iteration (total: 64, iterations: 32768)
RESULT : BenchHashes::typeRepo():"unordered_map":
0.0018 msecs per iteration (total: 59, iterations: 32768)

So QHash is just one magnitude slower than QVector, and still quite
fast... We might try that out in the future for future proofing the
duchain type registration.

Added slot that handles configuration changes specifically for WI settings dialog to avoid redrawing the entire skymap everytime WI settings is changed.

EclipsesPlugin: Update eclipses menu on config changes

KFileItemModelRolesUpdater: Avoid multiple updates for Nepomuk data

The Nepomuk ResourceWatcher emits 3 signals - propertyChanged,
propertyAdded and propertyRemoved. We should only listen to either the
propertyChanged signal or the propertyAdded + Removed signals. There is
no point in listening to all 3 signals. That will just result in
unnecessary updates.

Additionally, we do not need to listen to the resourceCreated signal.
That is only emitted when we are watching for a specific types, which we
are not.

REVIEW: 108543

Make initial inventory count depend on number of cores.

ResourceWatcherConnection: Do not emit propertyAdded/Removed signals

The information can be inferred from the propertyChanged signal. This
results in less dbus traffic.

BasicIndexingQueue: Use stacks instead of queues

Iterating depth first will consume less memory than going breadth first.
Also stacks are based on QVectors which avoid the unnecessary extra
allocation in QLists.

Also, free the QVector memory when it is no longer required.

Query Service Folder: Avoid multiple refreshes of queries

When the storage has changed, the query is re-run. If the storage is
continuously being changed then each time it is changed the query is
re-run (Unless the old query was still running).

>From now on, we use a timer and wait for 2 seconds after the storage has
changed in order to re-run the query. This has the negative impact of
not sending updates until there are no changes for over 2 seconds.

This might not happen in the case of indexing. However, this behaviour
is preferred over re-running the queries a large number of times.
Specially since the data mostly never changes.

FileWatch: ignore unnecessary ClosedAfterWrite signals

Whenever a file opened under write mode is closed, the ClosedAfterWrite
signal is emitted, even if the file has not been modified. This isn't a
problem as such cause the file indexer checks the mtime of the file
agains't the value in the database before indexing it.

However, as an optimization we can ignore all the ClosedAfterWrite
events, where the modification time of the file is less than 1 minute
from the current time.

This reduces the dbus traffic, and avoids checking the database each
time.

ActiveFileQueue: Immediately timeout when adding a file the first time

When a file is added the first time we instants timeout that file. If it
is added again within m_emittedTimeout seconds of the timeout, it is
then moved to the actual queue where it waits a certain amount of time
before it timeouts. If it is again queues during this wait, the timer is
reset.

The behaviour has been changed cause multiple enqueues of the same file
are the exception. We should not slow down the real-time indexing just
cause we could receive multiple events for the same file. With this
approach, both cases are suitably covered.

Tags kioslave: Use the Tag class internally

We had avoided using the Tag class for listDir cause that would mean n+1
queries when listing all the tags - One for fetching all their uris and
the other to fetch each tag's properties. We could instead just do it in
1 query by fetching all the uris and identifiers in one go.

However, this approach had the drawback that we would display the
nao:identifier, whereas all applications actually show the
nao:prefLabel. Also, we could not display the correct datetime of the
tags.

This approach is cleaner and uses the ResourceManager to handle all the
caching.

IndexCleaner: Simplify the queries a little bit

* Use FILTER NOT EXISTS instead of bif:exists
* Use abbreviated properties

IndexCleaner: Reduce the number of cleaning queries

The cleaner always checks for legacy data as well. This is not always
necessary as not everyone has legacy data. We have 2 kinds of legacy
data -

* configuration dependent
* independent

The independent data just needs to be removed once, after which it never
needs to be queried again. We check for this kind of data and switch on
a boolean config value when we're done. That way we never need to do it
again.

The "Configuration Dependent" data depends on which files are folders
have been marked to be indexed. For this kind of data we check if the
strigi indexGraphFor relation is present. If it is, then we must run the
queries. Otherwise, we ignore it.

VirtuosoInferenceModel: Do not inference rule creation

We only need to disable the nepomuk-ontology-group creation. Not the
inference rules. Those are required!

QueryService: Do not emit unnecessary entriesRemoved signals

Remember to always add the result in Folder::addResult to the
m_newResults. Earlier it was only added when the result was actually
"new", this resulted it sending entryRemoved signals for all resources
in the old query result.

Simplify distribution of selection-models.

Instead of relying on the proper call order and setting the selection model in the modelstack,
we replace each usage of the model by a getter, and create the model already properly connected on demand.

This way we not only get rid of call order problems, but we also simplify passing around all the models to the different views.
The new approach therefore simplifies adding new views.

Wrote a new FullMessageCombiner class that combines the header and body
parts of a message into one part.

Made Changes to the ComposerAttachments class to use the
FullMessageCombiner.

Added Assertions in the ctor of the ImapMessageAttachmentItem and
FullMessageCombiner.

Removed the headerPartPtr() and bodyPartPtr() methods from the
ImapMessageAttachmentItem.

Made the FullMessageCombiner::data() returns the bodyPart and headerPart
data directly without caching.

Delay akonadi and nepomuk lookups to compress them.

Introduce a per-addresseelineedit timer that delays completion
lookups via both akonadi and nepomuk directly by 500ms, so they
can be compressed. This avoids running too many queries when one types
quickly into an addresseelineedit, which helps with performance.

REVIEW: 108549

Speed up the pure-nepomuk address completion.

This avoids unions and regexps, like we do in the akonadi contact search
job now, thus speeding up the (large) queries considerably.

Reviewed and sparql-foo'd by vHanda and cmollekopf. Thanks, guys.

remove Sheets::Doc::updateBorderButton(), currently endless loop and view-, not doc-related

REVIEW: 108586

thanks zagge for review

Remove non-existing items from zorder listing in wdgdeformioptions.ui

--build warnings

Move Calligra::Sheets::ElapsedTime class to a private header that won't be installed

REVIEW: 107717

Patch by Philip Van Hoof, thanks

remove conflicting hardcoded names from not-code-referenced elements in .ui files (e.g. labels/layouts)

--build warnings

Avoid unneeded data copying in some loop code in Stage

Remove unneeded zorder from KWRunAroundProperties.ui (also had non-existing entries)

--build warnings

remove the last changetracking code from KoTextWriter

Remove non-existing buddy entries from .ui files

--build warningsw

Make sure to refresh monitor zone when subclip was selected:

Put kapptemplate parentmap files into one file.
Add argument parsing to postprocess.sh to easily postprocess only one subfolder of kdesdk.
Add kapptemplate1 branch for old kapptemplate code from before the rewrite.
TODO: make 802067 not add a bunch of removed files back into the tree...

R +10 -0 kdesdk/kapptemplate-parentmap [from: kdesdk/kapptemplate-parentmap-1 - 075% similarity]

Refactoring AmpacheAccountLogin

Cleanup code.
Better error detection.
Better error reporting.

common-kdesdk: don't try to match version branches with "__" prefix.

That part of the regexes never matches; there is never a double-underscore
prefixed branch that contains kdesdk. I think those CVS-era branches
may predate the creation of the kdesdk module!

The removal speeds up revlist generation and might speed up the actual
conversions slightly too.

Refactoring for Equalizer dialog

Add comments
Improve saving of pre-defined presets
Fix Can't edit & save an existing equalizer preset
Cleanup of the Equalizer stuff a bit, expecially the Equalizer Dialog



FIXED-IN: 2.8

Finish the conversion to resources for the ProgressWidget

...which was one of the biggest, now clean up and testing...

On selecting new game don't abort existing game without confirmation

REVIEW: 108460

Initialize some variables and check on destruction


REVIEW: 108604

Initialize some variables and check on destruction


REVIEW: 108604
FIXED-IN: 4.10

Consistent styles in tooltips


FIXED-IN: 4.6.0
REVIEW: 108492

Allow QtSharedMemory on win32-g++ builds

kde-projects: Don't filter out kde-build-metadata.

Specifically, when we have a non-master global branch, the new feature
to filter out module-set modules that don't have that branch was being
mis-applied to kde-build-metadata just to get some flags set right.

Instead, set the flags needed directly and skip the filter.

create the parser files before the lexer file, because this include it

Search the gfortran library before using it unconditionally

Hope it fixes cantor build in build.kde.org

SimpleIndexingJob: Refactor the code

Split it into a graph creation function which can theoretically be used
by others. This was initially done for some changes in the
nepomukindexer, which were eventually discarded.

Port KWin::errorMessage to new Xcb::Extensions

Code section had been in an #ifndef NDEBUG which is the reason why I did
not find the usage of Extensions there and why it always compiled
successfully.

Some data elements which got dropped needed to be added again like a name
for the extension and errorBase, etc.

Sorry for the inconvenience of a semi-broken master.

fix fileindexerconfig test

carefully handle the shortcut-circuit for || in c++

REVIEW: 108493

do not remove leading special chars

According to Mozilla docu only [^A-Za-z0-9_] is considered
a Word character - yeah :-(

Adjusting some of the default values in kwin.kcfg

* using the existing enum types
* adjusting the settings which recently changed.

Remove ZoomLevelInfo due to licencing issues.
And it is not yet decided if the slider will go in and how it will be
implemented. Removing for now

make layout a bit less hopelessy unbalanced

get rid of bold titles
use of bold test is in general discouraged and should be removed everywhere

Remove the mapping for kuiviewer

It has its own product with the exact name on bugs.kde.org, no need for
mapping to kdelibs|general

do not remove leading special chars

According to Mozilla docu only [^A-Za-z0-9_] is considered
a Word character - yeah :-(

Remove ThreadRunHelper.

Changes in the way Qt emits signals across threads have made it
obsolete.

change the size of the sub widgets a bit

Change brush shortcuts to [ and ]

To align with mypaint, gimp and photoshop

completely remove remaining change tracking code from KoTextLoader and continue to work on removing it from KoTextWriter

In Sheets remove some duplicated includes & put own header include first

Add documentation to a few common-rules files.

Notably it says what variables you're expected to set
before including the files, and what they mean.

Simplify recursion rules for cervisia.

I have verified this change doesn't affect the repository,
except for a "svn path=" in the commit message metadata.

RTL'izing text with a <span> does not work good enough. It will not end up right-aligning the text.
Use <div> to rtl texts again.
Initially, a <div> was used. However, it broke Renkoo and caused unwanted newlines.
Now we check and add the <div> only if the text is rtl.
Matches Adium's behaviour.

Fix build when TelepathyQt4 is in different prefix

REVIEW: 108602

better handling of new user accounts when loging in from ldap

not initialized members
code polish

Compiled smokegen for Qt 5.

Signed-off-by: Dimitar Dobrev