|
Thank you for posting, WM_CLOSE should be okay. You don't check the return value of PeekMessage (that should always be zero but its better to make sure...) and you should check Msg.hwnd whether its the main window hwnd or something else (it could be one of your dialog hwnds or any other HWND inside the IE process). I would write "if (PeekMessage()) { check for WM_CLOSE }". You can get the main window hwnd by performing the following operation on one of your activex hwnds that are placed directly/indirectly on your main window:
HWND GetRootWindow(HWND hwnd)
{
for (;;)
{
HWND parent = GetParent(hwnd);
if (!parent)
return hwnd;
hwnd = parent;
}
}
WM_SYSCOMMAND/SC_CLOSE is gui specific (it indicates that someone pressed X on the window), noone should put critical handling on this message as your window can easily get simulated WM_CLOSE messages without gui interaction/WM_SYSCOMMAND/SC_CLOSE.
|
|
|
|
|
Thanks again for your detailed reply. I will take care of this.
Manish Agarwal
manish.k.agarwal @ gmail DOT com
|
|
|
|
|
You are welcome! Ignore my previous post, I posted something here and the deleted it later but you probably received a mail notification! 
|
|
|
|
|
There is one message pump and one message handler per process, why would you want to duplicate the message pump of IE in a dialog box onclick handler?
Its uterly stupid.
|
|
|
|
|
For example because he may want dialogbox features on his dialog using IsDialogMessage() that wouldn't be done by the main loop of IE. Such dialogbox features include for example hotkeys like closing the dialog with ENTER/ESC, traversing the controls with TAB. Of course this own loop is a bit hacky but still can be a good balance between invested time and functionality. Sometimes its not you writing the code - it can be heritage from another guy who worked for the company years ago and your boss wont give you time to put together a better solution especially if the current one works in a more or less satisfactory way. The world is not perfect, and one day you will have to realize this.
|
|
|
|
|
You never heard of windows hooks then?
|
|
|
|
|
You still havent answered. Just because a dialog is interested in some messages and register for them why should it duplicate the message pump?
Thwere is one message pumop, one queue, one handler per process. Period.
|
|
|
|
|
The hook can indeed be a good solution. This still isn't a reason for you to talk to other people here using an arrogant style with degrading words. You can be the best programmer in the world if you are missing basic social skills and the respect towards others, noone wants people like you in his team. Respect is mutual or nonexistent.
|
|
|
|
|
Forget hooking, why shouold there be a duplicate message pump in the process? Dont you think it is stupid?
|
|
|
|
|
It is not stupid, but its an ugly workaround to a problem that has no fine solution. This own message loop workaround has several problems: it runs as long as our modeless dialogs are open and during this time it serves the messages to other windows as well, in this case if IE had its own modeless dialogs then dialog features wouldn't work on those dialogs as those are handled by IE's main message loops (same is true for other possible features that are implemented in the main loop of IE). We could probably find even more problems with this solution. The question is: how much time is available for the asker to finish this job and how much risk he takes if he chooses to implement dialog features for himself in his own dialogs by going with the message loop of IE. Sadly, programming for companies is often about saving time and playing safe and not taking risks for delivering correct solutions (unlike in hobby projects) especially if that would involve modifying legacy code that "still works". There are cases where the own loop is fine: for example the winapi MessageBox() does this (but its a modal dialog) and windows also running its own loop while you are moving a window by grabbing its title bar (and there can be other cases). There are probably many other cases where the own loop can be appropriate but a modeless dialog is definitely not among those.
|
|
|
|
|
You have already suggested using IsDialogMessage() so the dialog can register to receive required messages from the processes message handler. This is the suggested way of doing it so putting your own message pump in the dialog and peeking at messages IS stupid.
Any suggestion that bad code should be implemented OR maintained is also stupid. If it is crap, chuck it out.
Code is either 100% perfect or it is junk. Period. (I have spent 16 years in the WIndows kernel where this law is absoloutely true. ONE bug, and its a BSOD. Its either 100% perfect or unusable junk. )
|
|
|
|
|
Nothing is 100% perfect. Period. This is true even for drivers and we experience this at least monthly here (usb drivers for android, various drivers for 3d graphics). The size of driver code is usually much less than the legacy enterprise software codebase of some corporations so keeping drivers tidy is much easier than doing the same with some other software - especially because at a driver developer company you are probably expected to invest time in perfecting things while some other companies have to prioritize which part of the codebase to develop and you have time only to maintain the rest. It is usually not the programmer who decides about priorities.
|
|
|
|
|
pasztorpisti wrote: Nothing is 100% perfect.
So says someone who has never worked in the kernel...
pasztorpisti wrote: usb drivers
I didnt say all kernel engineers are perfect and the fact their crap code is so apparent to you is an indication that lack of perfection in kernel code is readilly apparent.
Everything else you say is wrong and inexcusable. Yes there are crap companies, crap code and crap engineers. However it is your choice whether to be one or not.
|
|
|
|
|
Erudite_Eric wrote: So says someone who has never worked in the kernel...
So says someone who thinks he knows everything... Are you sure about that statement? Maybe you are a mind reader, but I doubt that... You should stop insulting people and degrading others on the forum. You should search for a hacker forum that is usually full of people going with a style like yours. Maybe you are a good technician but at the same time you are like a 15 years old boy comparing his penis to that others'. This forum is here to help others with solutions not to offend and degrade each other. If you think you have a better solution to the problem than the ones already posted then submit yours without offending others. There is a difference between debate and offense/degrading words and you seemingly don't know where the line is between the two.
modified 11-Jul-13 8:52am.
|
|
|
|
|
So you have worked in the Windows kernel? What drivers did you write?
pasztorpisti wrote: This forum is here to help others with solutions
and not to give them stupid advice.
modified 11-Jul-13 10:01am.
|
|
|
|
|
I never worked with the windows kernel and wrote drivers only for linux for newly created hardware pieces several years ago. In my opinion programming kernel is just like programming anything else with a different api and a bit different rules and more pain in the ass. That's it. By the way, the kernel has nothing to do with the question into which we are posting this useless stuff.
And sorry but this is the end of our conversation. I'm visiting codeproject to spend a little of my free time with useful things like helping others and solve technical problems/debating about them. Comparing penises is not my business.
|
|
|
|
|
So you have no basis for saying 100% perfect is not acchievable since you havent been exposed to a system which demands it.
Which means your statement that non 100% code is OK because 'you havent got time to fix it' is wrong and bad advice.
And so if you dont like having it pointed out to you then dont offer people bad advice.
modified 11-Jul-13 12:09pm.
|
|
|
|
|
Hi,
I have a CView and some objects on it with a right click behaviour, using overwritten method OnRButtonDown(UINT nFlags, CPoint point).
But its not working on a windows8 tablet with touchdisplay (selecting object and keeping finger pressed for some seconds).
Doubleclick and stuff is working, just not the right mouseclick. But its working in standard comboboxes.
Are there any special events that I have to implement in order to catch the right mouseclick?
Thank you!
|
|
|
|
|
Did you enable WM_TOUCH messages by calling RegisterTouchWindow()?
Than the mapped mouse button messages are not generated.
|
|
|
|
|
Hey, thanks for the reply. This gives me a good starting point.
However, I cant seem to get it working.
I registered the touch handling via
if (!RegisterTouchWindow())
in the overwritten OnCreate method of CMDIFrameWnd
I defined the WM_TOUCH message:
#if(WINVER >= 0x0601)
ON_MESSAGE(WM_TOUCH, OnTouch)
#endif
(I checked the winver, it is correct.)
But I never get to the OnTouch method.
Any advice?
EDIT:
Interestingly it works through remote-desktop from the tablet but not as a fat client installation. No idea why though...
modified 9-Jul-13 2:53am.
|
|
|
|
|
It should be not necessary to use WM_TOUCH handlers. I just wanted to point out that the behaviour is as described when using them.
Did you test it on multiple touch devices? Maybe the reason is a non default system setting which would also explain why it is working remotely.
|
|
|
|
|
I dont have multiple touch devices (had to buy this one in order to be able to test).
Maybe i will have to test the parameters of the RegisterTouch...() method, although I am sure it wont make a difference.
Its good to know that those messages will be translated to normal mouse messages.
Thanks for your support so far!
|
|
|
|
|
Hi all,
I have a dialog based application in MFC using visual studio 2005.
After I close the application,
in CTestMainFrame::OnClose(), I am doing PostQuitMessage( 0);
With this the dialog box closes but my application still exists in Windows task manager.
Can you please help is there any other way I can kill my app completely.
Regards,
Sunil.
|
|
|
|
|
To close a modal dialog (the main dialog of a dialog based application is a modal dialog), call one of the EndDialog(), OnOK(), or OnCancel() member functions of CDialog. This will perform all necessary cleanup, close the dialog and return to InitInstance() of your application. The application is then terminated by returning FALSE from InitInstance().
|
|
|
|
|
Calling PostQuitMessage(0) should do the job but somewhere your application is buggy for sure, you have to check out all deinitialization code in your app. You put some buggy code to a destructor or a function called by mfc or by one of your own destructors during program shutdown that doesn't work as intended. Run the program in a debugger and if you can reproduce the bug in while debugging then try to check out what is goin on in your program by pressing the pause button in the debugger. Check out the callstack of all threads, especially that of the Main thread.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
