Issue 17

Bugfix time for the C# parser:
99.5% passed test cases is better than 95%.

There's a pretty tough corner case remaining, though.

If we get redirected when grabbing a file, tell the plugin about the final URL
and not the original one. Fixes playback of youtube videos embedded in other web sites.

Check the return value from dbus_connection_open_private().
Fixes KDED crashing on startup when D-BUS is unavailable.

If at all possible, this should go into 3.5.4.

Fix bug currently causing kde.org to crash.

The ScrollViewQt constructor uses "if (1 ||isFrameView())" to determine when
to construct a CanvasQt. CanvasQt assumes that the passed "ScrollView" object
is static castable to a "FrameView" -> as it's not, it'll crash!

The problem is that isFrameView() can't be accessed in the ctor yet, when
operating on FrameView objects as the FrameView ctor is not yet called.
(as detected by Zack (tm) )

Solution: Add a non-virtual init() function to ScrollView, call it from
the FrameQt object, afer constructing FrameView objects. Now isFrameView()
properly returns true, and the CanvasQt magically works fine ;-)

Please review, kde.org and test/crashOne.html work again w/o crashes on destruction.

If icons are dragged outside the desktop area, jerk them back in.
Fixes the long-hated issue with the desktop becoming scrollable.

Delete the receive buffer *before* sending the AT command, and not after, since this could delete incoming data if the phone is too fast.
First try of implementing an AT queue stack.

OASIS OpenDocument / MathML: Fix sequence loading and prepare for non text
support in token element's content (e.g. to support mglyph)

Much better fix for the 'ignore whitespace in oasis'. This makes the loading
follow the ODF spec, and where the spec is unclear, follow what OOo does.
Patch from "Benjamin K. Stuhl"

* Fix sorting
* Got it ok'd by the openusability people after minor rewording of the Fit* entries
* Implemented zoom menu entries after advice by the ou people in kivio

Collection Scanner resolves symlinks, what avoids duplicating files when there's links that point to other folders inside of the
collection. Please test, this might create problems.

proxy exceptions for musicbrainz lookups

Made sure in case of import, the number of imported bytes gets saved properly, so that average speeds are OK.

Fixed flickering when scrolling (which is all the time.)

Fix the preferences / config stuff.

RDP is now closer to working. That is, it still doesn't work.

Retire the "append underscore to last used nick after unsuccessfully
going through the identity's list of nicknames" behavior, which even-
tually runs into the NICKLEN wall. Instead, ask the user to enter a
new nick, or cancel the connection attempt on opt-out.
Patch by Raphael Kubo da Costa, and mentored by me.

Fix receiving of large YMSG packets.
Big packets that were splittet into several tcp packets were truncated and destroyed.
That fixes the problem of empty FileTransfer information in case of picture transfers. They have a small preview picture embedded which makes the
YMSG packet too big for one tcp packet.

In the recipe import dialog, keep identical recipe items in sync.

For example, the recipe "Candy Cane Cookies" may be under both the "Christmas" and "Cookie" categories. The recipe will therefore appear twice in the list of recipes
to import. Now unselecting the entry under one category also unselects every other entry as well.

It was possible before to unselect a recipe and it would still be imported because it existed selected somewhere else in the list... that was rather annoying, though subtle, and I'm glad it's fixed ;-) (that's another annoyance that couldn't be in 1.0).

1. Ordering in the completion-box is much better now. First the items are ordered by access-level, and then by inheritance-depth(see QObject ;)
2. Most of the Code-Completion-Options are read from CppCodeCompletionConfig now(I added the options there), and used in the code. CppCodeCompletionConfig is not yet filled from the GUI though.

A rather hackish solution to solve the difficult conflict
that made expressions like "a is SomeType ? if_exp : else_exp"
impossible. (Nullable expressions stole the question mark
from the conditional expression.)

As a result, 99.94% of my test case collection pass.
I believe that 6 of the remaining 11 files are invalid (research pending).
3 others are using an unspecified extension (which is supported by
MS and Mono though, so I'm going to implement it sooner or later).
One file shows that UTF-8 recognition is still not perfect,
and the last one is a virtually unsolvable very-last-corner case.

What I'm saying is that for now, I'm satisfied with the
parser's correctness, and now I'm going for real KDevelop code.
(C# part, here we come!)

1. Shut of the verbose-mode by default. That shoud speed up everything about the evaluation, but had caused problems in the past(but not anymore).
2. Add an addition level to the sorting of the completion-box: first the access-level, then the inheritance-depth, then the item-type.
3. Use the popup-box to show additional interesting information like the item's container-class, the item's type, source-file, and other interesting information that does not fit into the completion-box within the popup that was previously used exclusively for the comments. Comments are shown below now. This additionally solves the problem that emplty comments caused the wrong popup-box to stay, because every item has a comment now.

* Begin the Java language support part while jpetso
does the C# language support part! WOOHOO!

Begin the C# language part in parallel to manyoso's Java part.

And we're parsing java files... :)

All you need to start parsing java files is to change
the line PrimaryLanguage=C++ in the global project file
to PrimaryLanguage=Java.

Right now the codemodel is not hooked up and the AST is
not integrated.

I don't think CMake knows how to deal with .ll files, so
add the .cc file instead.

The kdevelop-pg loads the _G_contents char array in a weird way. Need
to change this so that it uses QByteArray or however the c++ parser
does it.

I commented out a few lines in java-io.cpp that were referencing main.cpp
these methods in main.cpp need to be moved into a separate file if we
want to keep them.

Next steps would be to integrate the AST so that it inherits KDevAST and then
we need to generate an actual codemodel that inherits KDevCodeModel. Once
this is done, we'll be able to see the codemodel in the codeview :)

Cheers,Adam

Add a new program that does some simple sanity checks on CMakeLists.txt files.
Usage:
cmakelint.pl CMakeLists.txt
or
find . -name CMakeLists.txt | xargs cmakelint.pl

Patch from Andreas Pakulat to start providing better Qt 4 support for KDevelop 3.4
This patch:

-provides a new tab on the C++ configuration for Qt options
-moves the include-style option to the new tab
-provides configuration options for qt version, qt directory and designer
integration on the new tab page
-removes the designer integration configuration from kdevelop's settings page
-uses the already existing qt root entry in the project dom to create paths for
designer and qmake
-makes the configured qt directory override the qt dir setting in trollprojectpart.cpp
but keeps an explicitly defined QTDIR environment variable
-changes the logic of qt version and include style so that the include style can be
chosen freely for qt4. only the qt3 style is available for qt3.

More to come, thanks for the patch!

- message-history works correctly
- user-message-tabs work correctly
- in-document-conversation uses the message-history now
- users are more consistent now
- more stuff I forgot

jpetso proudly presents:
...Automatic Error Recovery for kdevelop-pg!

Using this feature, a parser doesn't give up when it
encounters a parsing error. Instead, it proceeds to where
the rule could possibly go on after the recovered part,
and continues parsing from there.
(For parser junkies: it proceeds to the first found token
of the recovered item's FOLLOW set.)

Usage: (easy!)
Wrap your favorite optional items into recover(...).
Like, "(#stmt=statement)*" becomes "recover(#stmt=statement)*".

Note that the star should be outside of recover(...),
otherwise you'd only recover at the end of all statements
and the parser would not try for another statement.

Also contained in this commit:
* A fix for the FOLLOW set calculation for star and plus items
* Generating a few more newlines for better readability

a start for a matrix interface
added some apidox

Improvements to AltVsTime tool. I restored the crosshairs that appear when clicking or dragging the mouse to aid in measuring altitudes and times in the plot. In addition, the altitude and time of the crosshairs are now labeled on the plot. The time of the "Now" line is also labeled.

Moved kitenEdit to derive from KHistoryCombo (as Konq's bar does) for consistency and because doing it made it VERY easy to nick konqy code and make it display a searchbar. Next step for me: make it do something.

Added OpenGL selection code.

Single left click on an atom now selects/deselects it, if the mouse
didn't move between button press and button release. Some tolerance (2
pixels) added to make that usable with imprecise pointing devices.

TODO: notify the treeview (currently it doesn't get updated).

Change some things in the way OpenGL selection is done

It now happens on button press, instead of happening on button release. The clicked atom is highlighted in a separate color (currently some shade of purple, tell me what you think) even if it is not selected. The point of that is that this clicked atom is going to play a very important role in the mouse navigation system I plan to write. Any rotation/move will be done with respect to it. So I thought that this special role of this atom should correspond to a visual highlighting.

The "DST Rule" label now looks like an HTML link (the text is blue and underlined), and clicking on it will open the window explaining the available DST rules.

This window was a KMessageBox::informationList(), but I'm now creating my own KDialog, because I wanted to be able to increase the width of the window so there's no need for a horizontal scrollbar. Apparently the KMessageBox can't resize itself to fit the contents.

Begin work on making the calculator work. Most of the interactive moes work now, I haven't checked any batch modes yet.

I'd like to try to make the module layouts simpler and more intuitive.
Ideas are welcome!

digikam from trunk : improvement of camera interface : Restoring an old feature from digiKam 0.6.2 to upload pictures to camera.

In fact the interface to upload into gphoto2 camera have been never removed but the interface with the new digiKam folder view have never written by renchi.

All camera type are supported : Gphoto2, PTP, UMS, etc. This implementation have been tested with my Olympus C3000Z and my Minolta Dynax 5D.

By this commit, there is 2 ways to upload pictures in your camera :

- Drag digikam from trunk : improvement of camera interface : Restoring an old feature from digiKam 0.6.2 to upload pictures to camera.

In fact the interface to upload into gphoto2 camera have been never removed but the interface with the new digiKam folder view have never written by renchi.

All camera type are supported : Gphoto2, PTP, UMS, etc. This implementation have been tested with my Olympus C3000Z and my Minolta Dynax 5D.

By this commit, there is 2 ways to upload pictures in your camera :

- Drag & drop between digiKam main interface (or quesror if you want), and camera interface.
- Copy and Paste like between digiKam album.

Todo : add a new 'Upload' button on the bottom of Camera Interface. I need a feedback from user before to do it... This one is really mandatory ?

Note : with Gphoto2 camera, check if the camera dirver support upload using the new camera info dialog available by Camera Interface Help menu.

digikam from trunk: use same menu mnemonic for Filter as Krita and Photoshop

Add table of contents support for PDF files.

Add script from Rainer Krienke to export digiKam tags from database.

Support the user adding a new zoom level by typing in one and pressing enter. Zooming is now five by five again.

Allow entering of "123" zoom instead of only "123%". Previously, if it couldn't find a % to chop, it would kill the last digit (the '3' in "123", giving an incorrect, 12%).

Add hack to get around "10&0%" action text which sometimes happens with kdelibs4.

okular developers: after updating kdelibs, zooming should work perfectly - please let me know if it doesn't.

allow manually turning on of previews in the pager at low screen
resolutions
thanks to vinay shastry for testing the patch.

Add C# mimetype for KDE 4. (This is a temporary solution
until fd.o's XDG MIME-type database is supported.)

The "x-csharp" MIME-type is what XDG also uses
(see <a href="https://bugs.freedesktop.org/show_bug.cgi?id=231">https://bugs.freedesktop.org/show_bug.cgi?id=231</a>;)
and does not yet have a corresponding icon in the
Crystal or Oxygen iconsets.

There won't be a C# mimetype for KDE 3.x anymore.

Implement a (unique for toolkits?) feature suggested on
<a href="http://holehan.blogspot.com/2006/07/small-steps-to-fittser-kde.html">http://holehan.blogspot.com/2006/07/small-steps-to-fittser-kde.html</a>;

Makes all KStyle styled QLabels an additional clickable area which
sets keyboard focus on the corresponding buddy widget. Behaves exactly
like the given example, including mouse click visualisation using a
focus rect.

First stab at text/line edit rendering. Reviewed by Zack.

* Introduce new "Widget::setParentWidget" logic similar to "setWindowHandle" (windows backend) and/or "setDrawable" (gdk backend), which will create the actual widgets of ie. a LineEdit and associate the right parent with it.

Once a ie. <input type=password> object is constructed a new PlatformLineEdit will be created by the RenderText* objects, and ScrollView::addChild(Widget*) will be called. There we now know the parent widget and can properly construct our QLineEdit, as PlatformLineEdit::setParentWidget() actually handles the creation.

I hope this is flexible enough. Comments please :-)

* Remove parentQObject stuff from FrameQt (obsolete now)
* Adapt PlatformLineEdit/PlatformTextEdit and their *Qt implementations to that logic.

First step at making the fake backend useful:
1) capabilities are now exposed through D-Bus, so you can simulate button activation for example
2) the exposed computer is now much more complex and realistic

GUI: Some instructions and howto on the bluetooth pairing wizard.
Also, allow the user to cancel scanning, and added a button to resume services scan.
Some gui changes to make it look nicer.

progress - status right now is
-Simple shearing of a single non-rotated object works
-Simple shearing of mulitple non rotated objects works
-Shearing already sheared objects does not work (correctly)
-Shearing rotated object(s) does not work

A port of the siox tool from gimp / inkscape to krita. It is mainly based on inkscapes code, but uses some algorithms from gimp.
Unfortunately it is not finished yet. But I go on holiday tomorrow and I want to commit it before I leave.

Introduce the class UnderOverElement which will replace OverlineElement and UnderlineElement with a more general, more MathML compliant implementation.
It will support all kind of under and over scripts something we miss at the moment. This class will be used for munder, mover and munderover MathML tags.
APIDOX and functionality are yet to come.

reimport old olefilters for powerpoint as it gives better results than the new one

* Kexi supports database-aware combo boxes *
KexiDB - added support for db-aware lookup fields

Table View - added support for db-aware combobox fields
- displaying values in combo box popup fixed

2.0: merged

added a new command for changing the background of shapes

added command for aligning shapes

add infrastructure for showing context menus

add ATSC channel file and tuning support

This patch adds channel file and tuning for ATSC to kaffeine.
Note: Scanning isn't supported yet.

1) Use all ATF features that only require reading the UID at all times. GUI option now controls whether new UIDs are
written to files. Note that this helps generalize ATF; we can use this to now enable other writes to metadata if we
wish and have it be a global "yes you can touch my files" setting.
2) Only force a rescan/playlist clear the first time ATF is turned on (err, since this commit...)
3) If Amarok is closed before rescan finished, don't force it on restart...takes too long for some people.
4) ChangeLog updates.

"Some" work on tagdialog:
- Now we have statistics tab enabled when in multiple tags mode. Information there is pretty much a stub, I gotta improve that;
- It's now possible to edit scores/ratings for multiple tracks;
- Tag writing is done in a thread;
- Some clean up here and there, cutting some member variables we don't need anymore;
And of course, new bugs and regressions! Enjoy.

Lyrics can now be edited inside of context browser, no need to use tagdialog only for fixing some typos anymore.

sk if transfer to media device should be finished for all queued tracks or should be interrupted after current track when pressing disconnect

Add playlist tree, needs work for adding items (cp recent.xml
playlist.xml for a start). Had to figure out how to put an item/group in the active tree w/o the playlist tree starting to play all the way, solved by cloning the tree at item or group via outerXML and readXML.

New version of Score2Rating! (will hit kde-apps after packaging...)
* Better error handling, prints a list of failed tracks (together with my mail address) to STDERR when unsuccessful
(It prepares its own "backtrace", which ends up in Amarok's error dialog).
* Escapes all (?) bad characters which confuses the dcop query commands (now also including "double quotes").

love, skip love, skip & ban for last.fm streams in playlist right-click menu

Add DCOP call playlist: filenames which returns a list of the current filenames in the playlist. Might be more useful for some
people if it returned urls, but that should probably be a separate function, as this calls the MetaBundle's filename() method.
Thanks to Arash Abedinzadeh

Enable/disable prev/next buttons as playlist filter changes

Another step to support different backends.

Added resolution options for 1280x1024 and 1600x200, and changed the resolution names to better reflect the times.

Adding a bunch of new RDP keymaps.
Will backport this to 3.5 if we do a new release.

Add a "merge metacontact" action.

Patch by "Michael D. Stemle, Jr."
Thanks
<a href="http://lists.kde.org/?l=kopete-devel&m=115408872318115&w=2">http://lists.kde.org/?l=kopete-devel&m=115408872318115&w=2</a>;

Decode and store the embedded preview picture that is sent with all picture transfer invites.

The preview could be displayed in the filetransfer notification dialog.

new action icons for the ark program not sure the box is neaded at all

new action icons

new icons action/apps icons and fixing some names

The "Details..." button works now ;-)

In the Ingredient tab of recipe input, the "Details..." button now pops up a non-modal dialog that displays what information is missing for complete nutrient analysis.

The dialog is updated as ingredients are added and removed and is automatically closed if nutrient information is complete.

The status messages should be complete information-wise, but might be tweaked or re-worded for clarity and aesthetics (I do need to make the 'can't convert from any of..." message clearer).

initial import of a WorKflow command library for the Kate editor

Update ingredient weights to also include how the ingredient is prepared. CAUTION: For updating the database structure, it just wipes
out any existing weight info!

There is also a yellow status for approximations, and if a prep method is used, but there isn't a weight for it, the status will be
yellow and it will use an entry without a prep method. Note that the status will be red if there is missing info and approximations,
and will be yellow if there is no missing information (red status) but just approximations.

Most of the work is done, but it needs some tweaking (when editting, the prep method combo box is hidden unless you resize the Per
Amount column)

No further work has been done on loading USDA data, so imported USDA data still won't automatically use prep methods.

Got it!
Backgrounds from some image dir are now wokring.
(if you find the way to enable them. Ha-ha ;) )

When drawing plots, merge together segments of the curve that are roughly straight into one straight line before telling QPainter to draw it. Advantages:
- Painting is faster.
- Exporting to operation-based image formats (e.g. PDF, SVG) gives a smaller file size.
- QPaintEngines that can't handle small line segments (hello QPrintEngine!) are now happier.

Speed optimalisation.
Do not needlessly create a KoID and deep copy it as well as creating
empty strings for each of these calls.

Speed up ATF by using the entries in the database whenever possible instead of always re-reading the metadata.

Note: I have yet to verify that this code works as intended. I'll test ASAP, but I believe it should be correct. Hopefully I've
introduced no new bugs... :-)

Martin -- can you give it a go with your 6300-song playlist and let me know how much of a speedup (if any) is gained with this code?
Thanks!

Ok, I'm stupid. Why the HELL didn't I do the conversions this way earlier, it's about 500 times faster (seriosly!)
Anyway, here is Score2Rating v3.0, which is TOTALLY immune to strange file names, because it doesn't care about them. Can someone with PgSQL try this before I upload it to kde-apps? I've tested with MySQL and SQLite, so no need for that.

Make ATF a bit more efficient, and fix a stupid regression caused by the optimizations from two nights ago that would cause moved files to have their IDs scanned as empty.

Commit the patch partially.

Reduces time taken by HistoryLogger::readMessages(QDate) from 25 seconds to 0.5 seconds with a 200kb log file.

Abstract the backgroundparser so it can be shared
by multiple language parts. As mattr said, we don't
need no stinking separate background parser for every
piddling language part... Gosh!

- creating a branch for starting to work on qmake4 support for KDevelop3.4

create a directory for a linear algebra library

After consulting with kdevelop devels it's not feasible to provide a separate qmake4 build system, thus removing it again.

First big step in porting the PageView as subclass of the new shining QGraphicsView.
PageViewItem's are QGraphicsViewItem's, and are responsible for painting a page each. Similar thing for the rubber band selection, represented by a QGraphicsViewItem placed - when needed - over the pages.
The mouse modes are in the way of porting to the new view, and some have been improved.

The text selection now it's moved in an own mode (more logical); the text selections of a page are hold in the respective KPDFPage, and painted using PagePainter when the right flag is added.

What is left to do? The mouse drag for scrolling the view is not working, as well as the zoom-by-selected-area, that does not center the selected area. And - of course - clean the code and remove the old parts (when we are sure they won't be needed anymore).

Fixed also two really small bugs in part.cpp and area.cpp, I'll forwardport them soon.

Ported KPrinter to Qt4's painting system.
The main change is KPrinter's interception of QPainter::begin, QPainter::end
before they are passed to QPrint*. Before, QPaintDevice::cmd was used - but
since this no longer exists, KPrinter uses its own QPaintEngine and passes on
all the calls to QPrintEngine*. This is probably not the most elegant solution,
but it works.

yet more include directive cleanups.
I'm finding all these with a new krazy plugin. Be afraid :>

We won't anymore convert MathML to KFormula for loading,
we will directly parse MathML, so I defined the new interface for the loader and commented the old converter, who's code will just be altered a bit to fit the new task

Start work on new way to update podcastchannels, this approach should work with all feed's you can throw at it.
Remeber: podcastepisodes are never removed from the db, only the files are removed when purging.

Merge tvdocument and recents to common FileDocument base for reading and
writing XML, needed to remove tv setting support kmplayer < 0.9
Add source for lists, instead of abusing others, mainly for allowing
lazy loading files

MySQL v5 now required!

Update *every* dialog to be KDE-standards compliant. The biggest visible change is probably slightly different margins/spacing, but it uses KDialogBase to automatically apply standard behavior.

These are the changed dialogs:
-Ingredient Parser
-Category selection dialog
-Author selection dialog
-All dialogs for entering in a new ingredient, prep method, category, weight, etc.
-Shopping list (this was also updated to use the unit abbreviation setting)
-Diet results
-Rating editor
-Measurement converter
-Database import dialogs
-Resize recipe dialog
-Load USDA data dialog
-Border selection dialog
-And probably a few I forgot...

All dialogs now inherit KDialogBase instead of QWidget/KDialog/QDialog. KDialogBase automatically creates buttons like "OK" and "Cancel" and places them either along the bottom or right and also takes care of signals generated by these buttons. It adds other handy/standard behavior such as making "Esc" close the dialog, uses standard keyboard shortcuts, and follows the KDE setting of whether or not buttons display icons on them.
Basically, it adds standard KDE dialog behavior with no effort on my part and with less code ;-) (around 500 less lines total)