thomas@uvabick.UUCP (Thomas Fruin) (07/10/87)
Thought I'd try my luck on Usenet too, after not hearing from Info-Mac for about two weeks... I'm confused by a Window Manager problem I've run into. Maybe I've even found an error in Inside Macintosh. Let me explain what I'm trying to do: The program I'm working on will have a "palette" window similar to those windows in FullPaint: a window that always stays on top, but doesn't act as the active window. My program will have several document windows below this palette that should behave as normal as possible to the user. This means that when a user clicks in a window that isn't the active one, that window should be brought to the front, BUT stay below the palette window. Obviously I cannot use SelectWindow for this. Instead I use SendBehind, which seems exactly the function for this purpose (forget about highlighting for now). Here is where the problem occurs. Inside Macintosh tells me that if I use SendBehind to move a window closer to the front, I should make some extra low-level Window Manager calls after calling SendBehind. Specifically a call to PaintOne and one to CalcVis. But I don't think these are sufficient. Whenever I try this out by clicking in one of my windows to bring it to the "front", the visRgn of the window that used to be in front is screwed up. Anything I draw in it doesn't stay in its visible region, but flows into the window I just brought to the front. See the figure. It shows two windows before and after the SendBehind, PaintOne and CalcVis calls, when I start drawing into one of them. -------------------- -------------------- |behind | |in front | | ------------------- | |--- | |I am drawing text| | I am drawing text| ----|into lower right | -----into-lower-right | |window only. | |window only. | ------------------- ------------------- By changing the CalcVis call to a CalcVisBehind things go back to normal. So it seems to me SendBehind is not updating the visRgn of the window(s) it is covering up. Of course it should. The Apple traplist in the Sybex book Using the Mac Toolbox with C says that SendBehind already calls CalcVisBehind... Can anyone help me with this one? If I'm doing things all wrong, I'd also welcome suggestions to other and better ways to implement such a palette window. -- Thomas Fruin fruin@hlerul5.BITNET thomas@uvabick.UUCP Leiden University, Netherlands
oster@dewey.soe.berkeley.edu (David Phillip Oster) (07/13/87)
The current issue of MacTutor magazine has a good article on implementing pallette windows, and in particular it discusses the window manager issues involved. --- David Phillip Oster --My Good News: "I'm a perfectionist." Arpa: oster@dewey.soe.berkeley.edu --My Bad News: "I don't charge by the hour." Uucp: {seismo,decvax,...}!ucbvax!oster%dewey.soe.berkeley.edu
drc@dbase.UUCP (Dennis Cohen) (07/13/87)
The proper way to create palettes is via the "GhostWindow", not through the use of SendBehind, etc as described in your posting. Dennis Cohen Ashton-Tate Glendale Development Center dBASE Mac Development Team
gjs@k.cs.cmu.edu (Gregory Stein) (07/14/87)
I have implemented a palette with LSC. I found that keeping the window in the BACK rather than the front is painless and avoids many of the problems of trying to hide it in the front. The Toolbox has nothing against handling clicks in windows that aren't in the front. Just keep everything the same, but don't do a SelectWindow on the palette's window when you click in it. Also, since you'll probably be using a plain dialog box for the palette, you won't have worry about any hiliting effects. The code, depending on what it has to do, should not run you more than 100 lines, tops. To make it nice and flexible, you might want to read a list of PICT resources and position information from your resource file. --------- Greg Stein ARPA: gjs@k.cs.cmu.edu "There can be only one" -- from the movie, "The Highlander"
thomas@uvabick.UUCP (Thomas Fruin) (07/17/87)
There have been several replies and suggestions after my Window Manager query. Thanks to everyone. However, the question still stands. Let me first state that I _have_ implemented multiple palette/ghostwindows in Aztec C. Everything works fine. My code supports several palettes that always lie on top, except when a desk accessory gets opened or clicked. Then the disk accessory gets put on top. The "top" document window gets highlighted and activated normally. All windows can be dragged in their own plane, or are brought to the position just behind the last ghostwindow and then dragged. Several people suggested using the GhostWindow global. I've found it to be quite useless, because 1) its mechanism only supports ONE ghostwindow and 2) it does very little. Inside Macintosh says that GhostWindow can be made to point to a window that's not to be considered frontmost even if it is. Read those words carefully. It says nothing about what to do when you want to change the ordering of windows behind it. MacTutor's article about Palette windows is nice, but doesn't really help in this situation. It describes one popup window and only works when the windows below it don't change while the popup is visible. Thanks for the pointer though. Someone also mentioned that the low-level Window Manager calls I was using weren't preserving the current port. Fortunately I never ran into that problem. Gregory Stein writes: > I have implemented a palette with LSC. I found that keeping the > window in the BACK rather than the front is painless and avoids > many of the problems of trying to hide it in the front. The Toolbox > has nothing against handling clicks in windows that aren't in the > front. Just keep everything the same, but don't do a SelectWindow > on the palette's window when you click in it. Also, since you'll > probably be using a plain dialog box for the palette, you won't have > worry about any hiliting effects. We probably have very similar code. I don't call SelectWindow either to bring a window to the "front" (just behind my ghostwindows). This is where I use SendBehind. And being a good boy, I called PaintOne and CalcVis after- wards (see Inside Macintosh, I-286). This is where I think IM fails. It makes a genuine suggestion, but it is wrong here. Change CalcVis to CalcVisBehind and and everything works as it should. So this particular question remains: I am right in assuming this and is Inside Macintosh wrong? When I have my code cleaned up and commented, I'll be happy to post it. For now, see my modified SelectWindow. For simplicity I'm assuming there is only one ghostwindow, there are no desk accessories and every window is visible: /* Unhighlight and deactivate the current "front" window */ HiliteWindow( myGhostWindow->nextWindow, FALSE ); curDeactive = myGhostWindow->nextWindow ); /* Bring myWindow to "front", redraw it, recalculate its visible region and the regions of the windows behind it, highlight and activate it.*/ SendBehind ( myWindow, myGhostWindow ); PaintOne ( myWindow, myWindow->strucRgn ); CalcVisBehind( myWindow, myWindow->strucRgn ); HiliteWindow ( myWindow, TRUE ); curActivate = whichWindow; That's all folks! -- Thomas Fruin fruin@hlerul5.BITNET thomas@uvabick.UUCP Leiden University, Netherlands