Consider adding a system-sleep disabler

[Petr Janeček]

Hello!





When benchmarking serious (work) code I always try to run it on the hardware
it will eventually on, usually server-class machines.

However, when I benchmark code for fun (personal side-projects, code to 
prove somebody on the internet wrong...), I often run it

on my MacBook Pro. Please don't judge me, my company supplied it, I had no 
choice :).




When doing so, the following has proven very useful as a base class for my 
JMH-at-home benchmarks. It tries to disable screen

sleep, and, in effect, system sleep: https://gist.github.com/JanecekPetr/3c
9606dd6406c9c10672bd4b03eafa67





Would you consider adding similar functionality into JMH itself? It's 
annoying when I kick off a 30-minute benchmark only to come

back in an hour and find the benchmark was interrupted in half and the 
system went to sleep. 

Oh, and sometimes I need to subclass something else. Is there a way to use 
composition instead?




Thanks,

PJ








FAQ:

Q: WTH, why don't you disable system sleep in your system preferences?

A: Yeah I could, but I forget. And I do not want it to be completely 
disabled because I sometimes start benchmarks and go to sleep. Unneeded 
energy consumption and stuff.. And yes, I could set up a timer to shut the 
computer down after the benchmarks are done, but as I said, I forget.




Q: Should such a functionality even be in JMH?!

A: Not sure, you tell me! I think of JMH as a runtime for long-running 
tasks, and as such this functionality might be warranted, maybe?




Q: Argh, okay, maybe. Any surprises there?

A: Yeah, one *tiny* thing. On Mac (I only tested Windows and Mac) the 
program above triggers a Permission required notification for Terminal. 
Yikes. Interestingly enough, it still works if I deny the permission. Huh. 
But still, your Terminal probably should not ask for Accessibility 
permissions when running a JMH benchmark, so this should not be on by 
default. And if it is hidden behind a CLI option, I'll forget to use it...


Q: Running an extra timing thread with a background task sucks for very 
sensitive benchmarks.

A: Very sensitive benchmarks should not be running on systems with screen 
sleep. But yeah, I agree, maybe it'd be better to just wiggle the mouse 
between iterations if the iteration is expected to be shorter than a minute,
which it should almost always. I also tried using `Desktop.getDesktop().
addAppEventListener(new ScreenSleepListener() { ... });`, and the same with 
`SystemSleepListener` but at least on Mac using Java 17.0.1 this does not 
prevent the upcoming sleep. Weird.