bug in key repeat mode "7 Sticky"

[zuvutaba]

Version: 2.20.5
i use Installed version
Windows Version: Windows 10
Mouse Information: Warrior MO278
Computer Information: Xeon E5 2680v2 10 cores 20 threads 2.8Ghz turbo 3.6Ghz, 56GB RAM DDR3 1866MHZ
the problem occurred when I upgraded from version 2.19.2 to any 2.20.x including beta versions 2.21
I've always used it on Windows 10 I've never changed Windows
i use Xmouse for 2 years
keyboard PT/BR ABNT2

in the new versions when I use the repetition system

7 Sticky (repeatedly until button is pressed again)

when I'm using the repeat mode for about 3 minutes it starts to have delays to start the next cycle and this time it starts to take varies a lot sometimes it takes minutes for it to start the next repeat cycle

which does not occur in the older version 2.19.x, in the new version it consumes 10% of my CPU when in loop but in the old version it consumes a maximum of 2%,

I believe there is some error in the loop and that is why it is not able to start the next cycle correctly, I already reviewed my script, I created a simple one that presses only one key to make sure it was not my repeat sequence, I even went back to the old version to test and in the old version it works perfect, but the old version cannot use the scroll in firefox which disturbs me a lot

[phil]

I don't know about starting next cycle - I think you need to be a bit more precise, its not completely clear.
But the 2.20.4 and onward uses a new much more accurate timer, so if your repeat rate is < ~30 before it would never have gone faster than 30ms - now it can more accurately run, perhaps even to 1ms accuracy - of course most sequences take longer than that to run, so if that happens, it will skip and try again next time. That shouldn't make it take a long time though (unless maybe you have a massive waitms at the end of your sequence.

You didn't mention what the sequence is nor what your repeat rate is, so until I know that, I can not comment any further.
And I'm not saying its bug free, but I did do quite a bit of testing between 2.20.4 when it was first released and 2.20.5 - maybe something was missed. But more detail is required (the actual sequence) and a bit more about what you mean by "start the next cycle".

You could try 2.20.3 which does not have the new timer but should have the firefox fix, but ideally I'd like to know whats going on here to see if indeed it is something that needs attention/fixing, or if it is just a consequence of the new precise timer (in which case, increasing the repeat rate should help).

[zuvutaba]

when I say next cycle I mean the next repetition of the block of codes which in this case is

{END}{WAITMS:10}{LMB}{WAITMS:1000}{END}{WAITMS:10}{F12}{WAITMS:1000}

after 2 minutes of repeating this after the line ends it starts to take time to repeat again which does not occur in the old version

[phil]

Might be worth turning on debug logging, then let it repeat, and when it delays more, stop it and send me a copy of the debug log.
I havn't got much time to look into it at the moment but I have raised a bug report so I don't forget!

[zuvutaba]

hello, sorry for the delay in my return but I believe that what was happening is related to my hardware, because I recently changed a machine and did tests with a new Ryzen 9 5900x processor and what was happening on the old hardware does not happen, I did the test again on the old hardware and the bug happened again I think it is related to some incompatibility of my old hardware

[phil]

Interesting... It may be a timing issue rather than a hardware issue.
Trying to diagnose timing issues that are very hardware/software configuration specific is really difficult, especially when I can't make it go wrong here (I have a 5900x as my main dev machine)...

Keep an eye on this - if it is timing related, it may start happening again depending on what else if happening at the time.

[zuvutaba]

so when I use the "7 Sticky" repeat mode on my R9 5900x, I noticed that my cooler was accelerating so I decided to monitor its temperatures, with the game open it is around 49º ~ 51º, however when I start using the Xmouse with the sequence "{END}{WAITMS:50}{F5}{WAITMS:2000}", the processor temperature goes up to 62º ~ 65º but when I stop the repetition sequence the temperature goes back to 49º ~ 51º, what I I found it very strange for such a simple task to consume the CPU so that it heats up so much

another detail if I use the same sequence so that if I keep repeating it manually it does not use the CPU in this way it has a normal consumption nor does the temperature rise, only if I use the automatic repetition does this happen

what I noticed when using Xeon was that in these new versions with the new "7 Sticky" system and the high cpu usage

[phil]

Out of interest, what is your repeat time set to?

XMBC sets up a high precision timer (that's fairly new) which is quite CPU hungry but the only way to get precise, sub 30ms accuracy.

If your repeat time is less that the sequence time, it will be firing this timer every <repeat time> millisecond, processing the timer event, only to find it hasn't finished sending the sequence, thus do nothing but fire the timer again in <repeat time> ms. Basically one big CPU hungry loop which might explain the CPU usage & temperature.

As you have {WAITMS:2000} anything less that 2000ms repeat rate is surely going to suffer this, and the lower the repeat rate, the bigger the "problem".

Prior to 2.20.4 it didn't use a precise timer, it used a more simple timer that would not fire so fast and maybe therefore does not have such an impact IF you use a low repeat rate.

If your repeat rate is < 50ms try increasing it to 50, 100, 1000 and see if that makes any obvious difference?
NOTE: If its 0, it uses what ever Windows keyboard character repeat delay is set to for the first repeat, then whatever repeat rate is set to for subsequent repeats, which is slower for the first repeat cycle (about a second by default), then silly fast for subsequent repeats...

Untitled.png

[zuvutaba]

I removed the {WAITMS:2000} and put the "auto repeat time" of 2000 in the configuration and it really improved the situation, regarding CPU usage and temperature, thanks