*** Mon Oct 14 19:42:42 EEST 2002 - Dan Double buffering ---------------- To avoid flickering caused by redrawing the widgets on Expose events, a double buffering tehnique was implemented for most of the widgets. This flickering effect has gotten more vizible with the introduction of antialiased text. If with normal text one can redraw the text over the old one over and over again without any degradation of the text (new pixels simply overwrite old pixels), with antialiased text the situation is different. The pixels that constitute the antialias around the text are partially transparent pixels, which let part of the background be visible through them. If antialiased text is drawn over and over again, whithout first erasing the old one, the partially transparent pixels of the antialias will no longer see the background through them, but some of them will see the old pixels of the antialias around the old text that was there before. This for example will make a black antialiased text over a white background get thicker as the pixels of the antialias around it, combine with semitransparent black pixels of the old antialias instead of combining with the white background. The result is that the antialias will get darker (in this case) and the text will be altered. Because of this the area where text is drawn needs to be cleared before antialiased text can be drawn again. But between the moment whent he area is cleared and the moment when the text is drawn again there is a small time gap that results in flickering. This doesn't happen with normal text where one doesn't need to clear the area before drawing the text, but instead can simply draw over and over again. To avoid this situation, a double buffering tehnique was used. Instead of drawing directly in the wisget's window (which is mapped and will flicker as described above), we draw in a pixmap (unmapped) and after all is done we XCopyArea() from that pixmap to the real window. Since all this takes place off-screen, no flickering will be observed in this case. This is a change that that will be automatically available for your applications and will require no change from you. However there is an exception from this in case of WMList if you delegate the drawing of items to userspace (read below for the compelte details). *** Mon Oct 14 22:07:42 EEST 2002 - Dan WMList change ------------- In case of WMList there is the posibility to delegate the drawing of the list items to the application that is linked with WINGs, and this code will not be inside the WINGs library, but in userland. Since we use the double buffering tehnique in this case too (to allow all widgets based on WMList and the ones that draw their list items by themselves to benefit from the double buffering advantage automatically), we no longer pass the window to the user code doing item drawing, but instead pass this pixmap in which we draw before copying to the real window. Since one cannot use XClearWindow() or XClearArea() on pixmaps, but only on windows, if your list item drawing code used to contain these to clear the item area before drawing it needs to change, else the application will die when it tires to XClearArea() on a pixmap. This means that in your application if you ever used WMSetListUserDrawProc() so set a userspace routine which draws WMList items in a particular fashion, you need to make sure your function will not call XClearArea() or XClearWindow() on the drawable that was passed to draw in. Instead you should use XFillRectangle(). This change also means that you no longer need to do double buffering in your code, if you ever used to do. It is not done by WINGs and you benefit from it automatically. *** Mon Oct 14 19:28:35 EEST 2002 - Dan API change ---------- WMDrawString() and WMDrawImageString() no longer take a GC as argument. Instead WMDrawString() takes a WMColor* as the color for the string to be drawn, while WMDrawImageString() takes 2 WMColor* arguments in place of the old GC: first for text color and second for background color. This change is required to support extending WMFont to allow it to handle antialiased fonts through the XFree86 Xft extension. This also has the advantage of hiding low level X11 details and use WINGs internat objects instead. To fix your old code to work with the new WINGs API you need to replace the GC passed to WMDraw***String() in your code with a WMColor*. Most of the old code used to be like this: WMDrawString(screen, window, WMColorGC(color), font, x, y, txt, len); for the new API it should be replaced by: WMDrawString(screen, window, color, font, x, y, txt, len); However if you used a particular GC created by yourself to suit your special needs, you need to pass a color which is the same as the foreground color of that gc. For WMDrawImageString(), from: WMDrawImageString(screen, window, gc, font, x, y, txt, len); becomes WMDrawImageString(screen, window, textColor, backColor, font, x, y, txt, len); where textColor and backColor are declared like: WMColor *textColor, *backColor; and have the color of the foreground respective the background of the old gc. *** Wed Oct 9 07:10:04 EEST 2002 - Dan Xft support in WINGs -------------------- If Xft is detected when configure it is run, support for drawing antialiased fonts with transparency will be compiled into WINGs. You need at least Xfree86 version 4.0.x for this but at least 4.1.x is recommended. For Xft support there is a new function to create a font that will render using antialiasing and transparency: WMCreateAAFont(). Passing such a font to WMDrawString() or WMDrawImageString() will result in antialiased text displayed on screen. Modifying the alpha value for the WMColor passed to WMDrawString() or WMDrawImageString() will result in text being displayed with the appropriate transparency. To control antialiased font behavior, there are 3 new options that go into WMGLOBAL. Two of them are to set the system font used when an antialiased font is required. They operate in a similar way as SystemFont and BoldSystemFont, just they are used when antialiased fonts are requested. They are named AASystemFont respectively AABoldSystemFont. They are kept separate from SystemFont and BoldSystemFont because the same fonts don't render well as both normal and antialiased if they are not TrueType fonts. Even though you can specify any font in the XLFD format for these new options, it is recomended to use TrueType fonts for the antialiased case since other fonts (standard X fonts) don't render well and give ugly results. The third option is an option that globally controls if WINGs uses or not antialiased fonts. It is named AntialiasedText and it has a boolean Yes/No value. If set to Yes, WINGs will try to use antialiased fonts (if support was compiled in, and antialiased fonts can be found) then revert to normal fonts if not possible. Note that this applies if WMCreateFont(), WMSystemFont(), WMSystemFontOfSize(), WMBoldSystemFont() or WMBoldSystemFontOFize() are used. If any of the direct creation functions are used, such as WMCreateAAFont() or WMCreateNormalFont(), then that kind of font is returned independently of the value of AntialiasedText. (However note that an antialiased font cannot be created if Xft support was no compiled into WINGs, even if you call WMCreateAAFont() directly. In this case WMCreateAAFont() will always return NULL) *** Mon Sep 09 06:58:30 EEST 2002 - Dan New delegate for the WMConnection class --------------------------------------- ConnectionDelegate structure has a new member: canResumeSending. The purpose of this callback is to notify you that you can resume sending data over a WMConnection. It works in the following manner: WMSendConnectionData() can return 3 values: -1, 0, 1 -1 - means that the connection has died. you should stop sending data and close the connection ASAP. 1 - means that the data was succesfully sent 0 - means that the data (or part of it) was not sent. however, it was saved in a queue and the library will try to send it later when possible. if the return value is 1, you can continue to send the next message, and so on, until the return value of such a send call will be 0. In this case you can continue sending, however, the data will not be sent over the connection because the operating system cannot accept any more data for the moment. Instead it will be queued inside the library, making your program's memory footprint increase. If the ammount of data you need to send is limited and not too big, this shouldn't be a problem, because your data will be queued and sent when the operating system will notify the library that sending is possible again. If this is the case you can just ignore the output of WMSendConnectionData() and not set a callback for canResumeSending. However, if the ammount of data you have to send is undetermined and you also want to keep a small memory footprint for your program (so that it won't grow until it uses all your available memory ;) ), you will have to stop sending data over the connection as soon as WMSendConnectionData() returns with 0. Then you should somehow mark this situation in your program to avoid it trying to send anymore data until notified that it can resume. (You should have also set a canResumeSending callback when you initialized your WMConnection object because else you cannot be notified when to resume.) Now, when you receive such a 0 from the send operation, your last sent data is put in a queue inside the library. At a later time when the operating system notifies the library that sending is possible again, the library will resume to send the data that is saved in the queue. After it will be able to send all the data in the queue, the canResumeSending callback will be called, letting you know that not only you can resume sending because the operating system is again able to send data, but also that the queue was completely flushed. From the canResumeSending callback, you should again update the status of your program marking that it can send again, and then resume sending the data from where you were left. *** Thu Oct 04 06:00:09 EEST 2001 -Dan Property lists handling code ---------------------------- Code to handle property lists was added to WINGs. It is more robust than the libPropList code, mostly because some conflicting concepts borrowed from UserDefaults (which libPropList use) are no longer used in the WINGs property lists code. These borrowed concepts conflicted with the retain/release mechanism of property lists and could lead in certain cases to segmentation faults when executing libPropList based code. But the worse part was that these libPropList problems were practically unsolvable without removing one of those conflicting concepts and without a complete redesign. The new WINGs property lists code is also better integrated with the other data types from WINGs and is actively maintained. Practically the things that were removed from the WINGs property list implementation compared to the old libPropList implementation, are exactly the UserDefaults borrowed concepts that conflict with the retain/release mechanism: - The container of a proplist object and the associated functions are gone. - The filename associated with a proplist object and the corresponding functions are gone. Now the saving function needs the filename as a parameter. - The synchronization functions are no longer supported. They are part of the UserDefaults and are implemented there. - No functions related to domains/registering were implemented in the WINGs property lists code, because they are also not part of property lists. They are more in connection with UserDefaults and a central point of access for domains. The above 2 concepts: container and filename were added to libPropList just to let it support synchronization which was borrowed from UserDefaults. Property lists as defined in the openstep specification are just complex data structures composed of strings, data, arrays, dictionaries and a mix of them and are not associated with any file in particular. UserDefaults on the other hand are property lists read from a specific file and they associate that property list with that file and allow them to be synchronized. Old libPropList based code can still be used by linking against the WINGs library containing the new proplist code with minimal changes which are described in detail in the comments at the top of the WINGs/proplist-compat.h header file (the same file carries the #defines for mapping old libPropList functions to the new WINGs proplist functions). Our recommendation is to move to the new functions WINGs provide because they better integrate with other function naming conventions in WINGs. The proplist-compat.h header file is just a way to have old code up and running with minimal changes so that we can remove the old and unmaintained libPropList from systems while keeping to use old libPropList based code without rewriting it and it should not be used for other purposes. *** Sat Apr 21 09:12:09 EEST 2001 -Dan API change ---------- To allow a correct display of icon images with alpha blending in panels and other places where a WINGs based application may use them the following changes took place: 1. The following functions were renamed: - WMSetApplicationIconImage() --> WMSetApplicationIconPixmap() - WMGetApplicationIconImage() --> WMGetApplicationIconPixmap() - WMSetWindowMiniwindowImage() --> WMSetWindowMiniwindowPixmap() 2. The following functions were added: - WMSetApplicationIconImage(WMScreen *scr, RImage *image) - RImage* WMGetApplicationIconImage(WMScreen *scr) - WMPixmap* WMCreateApplicationIconBlendedPixmap(WMScreen *scr, RColor *col) As you can see the old functions that operated on WMPixmap images (which are basically X Pixmaps that lack alpha information) were renamed to ...Pixmap() to make them more suggestive about what they do and to make room for the new functions that operate on RImages (that hold alpha information). Since the corresponding WMGet... functions only retrieve the stored image/pixmap from the application, I'll outline how the WMSet... functions operate: All WM...IconPixmap() functions operate on WMPixmaps All WM...IconImage() functions operate on RImages - WMSetApplicationIconImage() will set the RImage to be used in panels and will also convert the RImage to a WMPixmap with a threshold of 128 and will use that pixmap for the appicon image. If that doesn't satisfy you, you can make a call to WMSetApplicationIconPixmap() on your own to set whatever WMPixmap you see fit for the appicon. - WMSetApplicationIconPixmap() will set the WMPixmap to be used for the appicon and for the panels If you use only one of the above functions, the corresponding image/pixmap will be used everywhere where needed (panels and appicon), but the pixmap version will not be able to handle alpha blending correctly. If you use both WMSetApplicationIconImage() and WMSetApplicationIconPixmap() then the RImage will have priority in panels, and the WMPixmap will only be used for the appicon. This allows you to better control what icon is displayed in the appicon, in case the default conversion of the RImage to a pixmap with a threshold of 128 is not good enough, or in case you want a different icon to be shown in the appicon than in panels. Also this new function was added: - WMCreateApplicationIconBlendedPixmap() will use the RImage set with WMSetApplicationIconImage() if available and will blend it with the color you passed. This will make the image show well on a background of that color. If the RImage was not set it will return NULL. You need to call WMReleasePixmap() on it after you finish with it. Passing a NULL pointer instead of a color will make the RImage be blended with the default color of the WINGs widgets: '#aeaaae' making it suitable to be assigned to any WINGs widget. To make your existing code work as before all you need to do is to rename the following functions: - WMSetApplicationIconImage() --> WMSetApplicationIconPixmap() - WMGetApplicationIconImage() --> WMGetApplicationIconPixmap() - WMSetWindowMiniwindowImage() --> WMSetWindowMiniwindowPixmap() But if you want to take advantage of the new abilities to show alpha blended images you need to start using the new functions.