Home | All Classes | Main Classes | Annotated | Grouped Classes | Functions

The Qt/Embedded-specific classes

Qt/Embedded classes fall into two groups: the majority are used by every Qt/Embedded program, and some are used only by the Qt/Embedded server. The Qt/Embedded server program can also be a client, as in the case of a single-process installation. All Qt/Embedded specific source files live in src/kernel and are suffixed _qws. The » symbol indicates inheritance.


There is one of these per application. At application startup time it reads the font definition file from $QTDIR/etc/fonts/fontdir (or /usr/local/etc/qt-embedded/fonts/fontdir if QTDIR is undefined). It keeps track of all font information and maintains a cache of rendered fonts. It also creates the font factories: QFontManager::QFontManager is the place to add constructors for new factories. It provides a high-level interface for requesting a particular font and calls QFontFactories to load fonts from disk on demand. Note that this only applies to BDF and TrueType fonts; Qt/Embedded's optimised .qpf font file format bypasses the QFontManager mechanism altogether.

There should be no need to modify this class unless you wish to change font matching or caching behaviour.


This contains information about a single on-disk font file (e.g. /usr/local/etc/qt-embedded/times.ttf). It holds the file path, information about whether the font is scalable, its weight, size, Qt/Embedded name, etc. This information is used so that QFontManager can find the closest matching disk font (it uses a scoring mechanism weighted towards matching names, then whether the font's italic, then its weight).

There should be no reason to modify this class.


There is one and only one QRenderedFont for every unique font currently loaded by the system (that is, each unique combination of name, size, weight, italic or not, anti-aliased or not). QRenderedFonts are reference counted; once no one is using the QRenderedFont it is deleted along with its cache of glyph bitmaps. The QDiskFont it was loaded from remains opened by its QFontFactory.

There should be no reason to modify this class, unless you wish to change the way in which glyphs are cached.

QFontFactory (and descendants QFontFactoryBDF, QFontFactoryTtf)

These provide support for particular font formats, for instance the scalable Truetype and Type1 formats (both supported in QFontFactoryTtf, which uses Freetype 2) and the bitmap BDF format used by X. It's called to open an on-disk font; once a font is opened it remains opened so that the creation of new font instances from the disk font is fast. It can also create a QRenderedFont and convert from Unicode values to an index into the font file. For compactness, glyphs are stored in the order and indexes they are defined in the font rather than in Unicode order.

There should be no need to modify this class, but it should be inherited if you wish to add a different type of font renderer (e.g. for a custom vector font format).


This describes a particular image of a character from a QRenderedFont: for example, the letter 'A' at 10 points in Times New Roman, bold italic, anti-aliased. It contains pointers to a QGlyphMetrics structure with information about the character and to the raw data for the glyph: this is either a 1-bit mask or an 8-bit alpha channel. Each QRenderedFont creates these on demand and caches them once created (note that this is not currently implemented for TrueType fonts).

You would only need to modify this class if you were, for example, modifying Qt/Embedded to support textured fonts, in which case you would also need to modify QGfxRaster.


This handles requests for space for pixmaps and also keeps track of QPF format fonts (these are small 'state dumps' of QRenderedFonts, typically 2-20KB in size; they can be mmap'd direct from disk in order to save memory). If a QPF font is found which matches a font request no new QRenderedFont need be created for it. It's possible to strip out all QFontFactory support and simply use QPFs if your font needs are modest (for instance, if you only require a few fixed point sizes). Note that no best-match loading is performed with QPFs, as opposed to those loaded via QFontManager, so if you don't have the correct QPF for a point size, text in that size will simply not be displayed.

There should be no need to modify this class.

QScreen » QLinuxFbScreen » accelerated screens, QTransformedScreen » QVfbScreen

These encapsulate the framebuffer Qt/Embedded is drawing to, provide support for mapping of coordinates for rotating framebuffers, allow manipulation of the colour palette and provide access to offscreen graphics memory for devices with separate framebuffer memories.

This is used for caching pixmaps and allowing accelerated pixmap=>screen blt's. QLinuxFbScreen and the accelerated screens use the Linux /dev/fb interface to get access to graphics memory and information about the characteristics of the device. The framebuffer device to open is specified by QWS_DISPLAY. Only QTransformedScreen implements the support for rotated framebuffers. QVfbScreen provides an X window containing an emulated framebuffer (a chunk of shared memory is set aside as the 'framebuffer' and blt'd into the X window): this is intended as a debugging device allowing users to debug their applications under Qt/Embedded without leaving X. The accelerated screen drivers check to see if they can drive the device specified by QWS_CARD_SLOT (which defaults to the usual position of an AGP slot if not specified) and mmap its on-chip registers from /dev/mem. They may also do chip-specific setup (initialising registers to known values and so on). Finally, QScreen's are used to create new QScreenCursors and QGfxes.

If you wish to modify the way pixmaps are allocated in memory, subclass or modify QLinuxFbScreen. If you're writing an accelerated driver you will need to subclass QScreen or QLinuxFbScreen.

QScreenCursor » accelerated cursor » QVfbCursor

This handles drawing the on-screen mouse cursor, and saving and restoring the screen under it for the non-accelerated cursor types.

Subclassing QScreenCursor is optional in an accelerated driver (you would only want to do so if the hardware supports a hardware cursor).

QGfx » RasterBase » Raster » accelerated driver » QGfxVfb » QGfxTransformedRaster

This class encapsulates drawing operations, a little like a low-level QPainter. QGfxRaster and its descendants are specifically intended for drawing into a raw framebuffer. They can take an offset for drawing operations and a clipping region in order to support drawing into windows. You will need to subclass the QGfxRaster template in order to implement an accelerated driver.

If you're brave, modifying QGfxRaster would allow you to customise how drawing is done or add support for a new bit depth/pixel format.

QLock, QLockHolder

This encapsulates a System V semaphore, used for synchronising access to memory shared between Qt/Embedded clients. QLockHolder is a utility class to make managing and destroying QLocks easier.

There should be no need to modify this class unless porting Qt/Embedded to an operating system without System V IPC.


This is a QPainter which also gives you a pointer to the framebuffer of the window it's pointing to, the window's clip region and so on. It's intended to easily allow you to do your own pixel-level manipulation of window contents.

There should be no reason to modify this class.

QWSSoundServer, Client

The Qt/Embedded server contains a simple sound player and mixer. Clients can request the server play sounds specified as files.

There should be no need to modify this class unless porting Qt/Embedded to an operating system without a Linux-style /dev/dsp.


This contains the server's notion of an individual top level window: the region of the framebuffer it's allocated, the client that created it and so forth.

There should be no reason to modify this class.

QWSKeyboardHandler » subtypes

This handles keyboard/button input. QWSKeyboardHandler is subclassed to provide for reading /dev/tty, an arbitrary low-level USB event device (for USB keyboards) and some PDA button devices.

Modifying QWSKeyboardHandler would allow you to support different types of keyboard (currently only a fairly standard US PC style keyboard is supported); subclassing it is the preferred way to handle non-pointer input devices.

QWSMouseHandler » QWSCalibratedMouseHandler » mouse types

This handles mouse/touch-panel input. Descendants of QWSCalibratedMouseHandler make use of filtering code which prevents 'jittering' of the pointer on touchscreens; some embedded devices do this filtering in the kernel in which case the driver doesn't need to inherit from QWSCalibratedMouseHandler.

Subclassing QWSCalibratedMouseHandler is preferred for touch-panels without kernel filtering; inheriting QWSMouseHandler is the way to add any other type of pointing device (pen tablets, touchscreens, mice, trackballs and so forth).


This class exists only in the Qt/Embedded server and keeps track of all the top-level windows in the system, as well as the keyboard and mouse.

You would only want to modify this if making deep and drastic modifications to Qt/Embedded window behaviour (alpha blended windows for example).


This manages the Qt/Embedded server's Unix-domain socket connections to clients. It sends and receives QWS protocol events and calls QWSDisplay in order to do such things as change the allocation region of windows.

The only reason to modify this would be to use something other than some sort of socket-like mechanism to communicate between Qt/Embedded applications (in which case modify QWSClient too). If you have something like Unix domain sockets, modify QWSSocket/QWSServerSocket instead. Don't add extra QWS events to communicate between applications, use QCOP instead.


This encapsulates the client side of a Qt/Embedded connection and can marshal and demarshal events.

There should be no reason to modify this except to use something radically different from Unix domain sockets to communicate between Qt/Embedded applications.


This manages a client's QWSClient, reading and interpreting events from the QWS server. It connects to the QWS server on application startup, getting information about the framebuffer and creating the memory manager. Other information about the framebuffer comes directly from /dev/fb in QLinuxFbScreen.

There should be no reason to modify this.


These encapsulate the data sent to and from the QWS server.

There should be no reason to modify them.


QCop is a simple IPC mechanism for communication between Qt/Embedded applications. String messages with optional binary data can be sent to different channels.

The mechanism itself is designed to be bare-bones in order for users to build whatever mechanism they like on top of it.


This provides Qt/Embedded window management, drawing a title bar and handling user requests to move, resize the window and so on.

There should be no reason to modify it but you should subclass it if you want to modify window behaviour (point to click versus focus follows mouse, for instance).


Descendants of this class are different styles for the Qt/Embedded window manager, for instance QWSWindowsDecoration draws Qt/Embedded window frames in the style of Windows CE.

Subclass it in order to provide a new window manager appearance (the equivalent of a Windows XP or Enlightenment theme).


This provides the QWS client's interface to the QWS property system (a simpler version of the X property system, it allows you to attach arbitrary data to top-level windows, keyed by an integer).

There should be no reason to modify it.


Used by both client and server to help manage top-level window regions.

There should be no reason to modify it.

QWSSocket, QWSServerSocket

Provides Unix-domain sockets.

Modify this if you're porting to a non-Unix OS but have something analogous to Unix-domain sockets (a byte-oriented, reliable, ordered transmission mechanism, although you can probably implement it with something like a message queue as well).

Copyright © 2005 TrolltechTrademarks
Qt 3.3.7