[Quiz] What will happen if SendMessage to self?

[myron]

This is an interesting thing (what i never thought of yet...) and I wonder how you people think it will happen.

We normally won't sendmessage to ourself, but what if we do that?

Here are the sample example:

LRESULT CDlgExeDlg::DefWindowProc(UINT message, WPARAM wParam, LPARAM lParam) 
{
	// TODO: Add your specialized code here and/or call the base class
	switch (message)
	{
	case WM_MYTEST:
		{
			if (! SendMessage(WM_MYTEST)) {
				ASSERT(0);
			};
			
			int i = 0;

		}
.....

Because SendMessage is a synchronous function, so THEORETICALLY it will casue an infinite loop!

But is this true?

How do you people think?

[VladimirF]

...Because SendMessage is a synchronous function, so THEORETICALLY it will casue an infinite loop!

Yes it will, and then - stack overlow...
Did you try?

[myron]

m. but actually (usiang VC6)

it just runs!
(After a certain number of stack-record repetition...)
A very interesting things...

Says, what if we repalce WM_MYTEST with WM_CREATE?

The dialog/app still shows up!

What a magic.

[Martin O]

We normally won't sendmessage to ourself, but what if we do that?

We normally don't send messages recursively to ourself (i.e. send WM_MYMESSAGE inside a WM_MYMESSAGE handler), but other than that its very normal to send messages to ourself.

[myron]

Well, there is a trcik here.
Sendmessage is synchronous function, so "theoretically" this will cause system hang out... since the sendmessage won't return back... (in single thread condition of course)

...
case WM_MYTEST:
{
if (! SendMessage(WM_SOMEOTHERMSG)) {
ASSERT(0);
};
...

[NMTop40]

it's the first sign of madness

[Martin O]

Well, there is a trcik here.
Sendmessage is synchronous function, so "theoretically" this will cause system hang out... since the sendmessage won't return back... (in single thread condition of course)

Unless I'm mistaken it won't necesarily hang...first execution is in the case WM_MYTEST code, then when you call SendMessage(WM_SOMETHINGELSE), execution goes to the WM_SOMETHINGELSE handler. When that handler completes, execution returns to the WM_MYTEST handler, after the SendMessage call.

Now, if we're talking about Windows messages, not user defined messages, it may or may not cause problems, depending on the message & what Windows does with it.