GDI leak driving me nuts

[Greenhorn]

And b), your dialog proc is supposed to exit by calling EndDialog(), not by calling PostQuitMessage().

No, that's not correct, at least by empirical evidence. Perhaps it's a special case, since my dialog is the main window (i.e., I'm not invoking the dialog from another window but from my WinMain() procedure), but EndDialog() will indeed end the dialog but not the program, which remains stranded until closed by Task Manager. So PostQuitMessage() is indeed the correct thing here. It's what I've been doing in all my programs that are built as dialog-as-main-window since forever.

Japheth's statement is correct. I have several applications which use a Dialog Box as the main window and no problems with exiting the process.
Just call EndDialog, hWin, TRUE and in WinMain return the return value of DialogBox... function.
I trtied it in your application (and removed the useless message pump) and it works properly. 

[NoCforMe]

Sometimes one does not see the forest for the trees ... 

INVOKE    EndPaint, hWin, ADDR ps

That was it. Bingo.

You, esteemed coder, win the grand debugging prize. (As well as the careful-reading prize.) The GDI object count climbs up just a little bit (47 in my trial) and stays there.

My hat is off to you!

[NoCforMe]

I want to thank the congregation here for helping me and for solving this problem. I learned two things:

1. The message loop I have in all my dialog-as-main-window programs is totally unnecessary. I'll be getting rid of them.

2. Don't miss the forest for the trees.

Thanks again!

[jj2007]

Congrats, Greenhorn - you spotted what nobody else saw

I had a suspicion that it was the DC - see line 62 of the attached code above: there was no leak with CS_OWNDC (meaning Windows reused it all the time, and didn't mind that it was recreated a hundred times with BeginPaint).