I've started to encounter a strange problem when I try to quit my program. In 8 of 10 tries, the program still runs (as seen in my attached screenshot) and I have to terminate it with force (ctrl-alt-del). I don't think there's anything wrong with my code since I've seen this happen with other allegro programs, and the last thing my program does is writing to a logfile which it accomplishes.

I'm using the latest WIP and Windows XP.

Any idea what might be wrong?
This is getting kinda annoying.

99% chance there is a memory leak in your code.

do any of the allegro examples hang on exit ?

I have the same problem with my programs on Win XP.

All using the latest WIP?
And reproducible with the Allegro examples/test programme/demo game?

I ran you attached program ten times and it quits cleanly with ESC.
I'm on Windows XP SP1 (+ some updates)
I never had this kind of problems.

Just an idea: which version of DirectX are you using?

I have the same thing. With MSVC++ 6.0 I'll press escape (to exit the program) and the window closes, but it's still running. I have to do a shift-F5 to stop the debugger. I know I have memory leaks but why would that effect closing the program.

When you have memory leaks and you use a debug build, then MSVC runtime tries to collect infos about the allocated but not free'd memopry chunks. It's that what needs time, so that's probably what you have here.

If it is this it doesn't depend on the allegro version or any library, it always occurs if you have a debug build and a memory leak, so the runtime tries to find out all the details.

I have that problem without debugger. Allegro quiting functions should be checked.

I've done some testing and I think I am on to something.

I changed my program so that it didn't rely on allegro's atexit routines, I manually placed all the remove_timer(), remove_keyboatd(), yada yada, functions and recompiled.

I ran it 50 times and it worked like a charm.

Then I changed it all back, recompiled. I ran it 7 times until it collapsed on exit again.

THEN I changed it back AGAIN

I ran it 39 times, but this time it froze upon exit but not like usual. It seemed like the program was still running, but unable to get input.

But here comes the interesting part. After I quit the program I looked through my logfile. I had logged messages before almost every function and now it got to right before remove_timer() but didn't log the message right after that one.

Conclusion, the program froze when running the remove_timer() function.

Now, what to do with this information is up to someone else

I used the wrong dll:s. I have placed some logs in remove_timer() and will see if removing #undef... is going to solve it.

EDIT: Nope, that wasn't the solution.

The logfile indicates that the error happens in timer_driver->exit()

EDIT2: I went further and added logs in tim_win32_exit()from wtimer.c

The logfile indicates that the error happens in WaitForSingleObject(timer_thread, INFINITE)

After reading about WaitForSingleObject() on msdn, I understand that this is where the program gets stuck. Why not use a specified interval? If the return value is WAIT_TIMEOUT try to close the handle anyway. That's what I have to do manually anyway

EDIT3: I changed tim_win32_exit() to wait 5 seconds and then quit anyway. Worked like a charm, although I don't know what malice comes out of it.

static void tim_win32_exit(void)
{
   /* acknowledge that the thread should stop */
   SetEvent(timer_stop_event);

   /* wait up to 5 seconds for thread to end */
   if (WaitForSingleObject(timer_thread, 5000) == WAIT_TIMEOUT)
       al_trace(" wtimer.c: thread didn't end\n");

   /* thread has (hopefully) ended, now we can release all resources */
   CloseHandle(timer_stop_event);
}

Now, I'm gonna revert my program code to using allegro's atexit stuff. Hopefully, the thingy on the taskbar will dissappear after 5 seconds.

EDIT4: YESSSS! That was indeed the problem! The program quits after 5 seconds and no more thingys left on the taskbar. Sweet...

What do the developers have to say about this? What should be done?

Something is wrong if the timer thread isn't being killed properly..