AutoHotkey Community

I believe that if you use this after each joyConfigChanged(), it will decrease memory usage, not sure

_________________
fincs
Highly recommended: AutoHotkey_L (see why) (all my code snippets require it)
Formal request to polyethene - I support the unity of the AutoHotkey community
Get SciTE4AutoHotkey v3.0.00 (Release Candidate)
[My project list]

Intriguing idea! I modified the test script accordingly (with displaying the driver and the current time, so un/plugging the joystick is seen, and a crash is apparent). The memory usage now jumps up and down. It still increases, albeit much more slowly. After 10 minutes it increased from 640KB to only 1MB, instead of the earlier 200MB. Since the timer runs 100 times slower in the real script, it might not crash in a day or two with this modification. Still, eliminating the memory leak would be nicer.

Laszlo, if you get a chance please try tonne's idea of VMMap that I mentioned in my previous post. VMMap will hopefully give a more specific location for the where the extra memory is being allocated.

Unfortunatelly, the script in my previous post with fincs' idea (DllCall("SetProcessWorkingSetSize"... ) does not work. After about 50 minutes the PC crashes. I tried it three times, with the same results.

VMmap killed TaskManager, and did not provide any info I understand. I need the help of a Windows expert...

This exhibits the same behaviour:
A full equivalent to JoyName:=GetKeyState("JoyName") would be:

Thanks, Lexikos! I tried joyGetDevCapsA (as you saw in my Ask for Help post), and JoyGetPos, but after a joystick is unplugged several minutes can pass until the result becomes 167 (JOYERR_UNPLUGGED). I gave up waiting, thinking it did not detect unplugging. Do you know a way to ask Windows to rescan the joysticks? Maybe once or twice a minute. 5 minutes delay looks incredibly long.

Actually, I had not seen your Ask for Help post. I merely looked into the AutoHotkey source and found that GetKeyState("JoyName") uses joyGetDevCaps. I tested joyGetPos with my wireless Xbox 360 controller, and the return value immediately reflects disconnection/reconnection. joyGetPos is the recommended method according to MSDN as quoted in my previous post.

Hello

I try to UnZip the compiler file and i end up with a big bin file. What am i doing wrong ?

You need to drop it in the following folder
?:\Program Files\AutoHotkey\Compiler

.. and you need to recompile your existing compiled scripts to reflect the upgrade.

Hello

AAAh Ok i don't have the compiler. Can you just give us the exe ?

Sure!

http://www.autohotkey.com/misc/AutoHotk ... 8-new4.exe

The following script crashes after a few minutes, with a little, slow, but constant increase of memory footprint:If I don't use dynamaic variables the script looks more stable.

I'm still hampered by the lack of a joystick (all I have is an ancient one that needs a joystick port I no longer have). But I ran it in the debugger for about 15 minutes and nothing unusal happened. The memory increased very slowly, perhaps 4 KB every 5 minutes.

If you ever get bad behavior with routines other than joystick, please let me know. In the meantime, maybe others can test your script above in case it's an issue with one particular kind of joystick or driver.

Since the script does not create or enlarge variables after the first second, is not even a slow memory increase unusual? If I remove the joystick calls, the memory footprint remains constant.

Yes, seeing memory usage slowly increase would be cause for concern. However, the memory seems to be increasing in the shared working set, which the program doesn't do anything with, and maybe has no direct control over.

Since the private memory ("VM Size" in Task Manager) isn't increasing, perhaps that proves that the program is not allocating any memory from the stack or heap. This is supported by those unreached breakpoints I did for your prior script, though I haven't done it for this one.

VMMap seemed quite promising because it revealed exactly what DLLs or modules are using memory. But on my system, the memory increase was so little and took so long that VMMap probably wouldn't give a definitive answer.

If no one else with a joystick is able to get useful results from VMMap, I can try running a script overnight again to see what it shows.