Process::cpu_use() is off on Linux about sysinfo HOT 9 CLOSED

Try to run System::refresh_all at least once. If it works after this, then it means some values aren't initialized correctly on sysinfo side.

from sysinfo.

I released the new version.

from sysinfo.

Cool, thanks for doing it so quickly :).

from sysinfo.

Code looks correct so that's weird... Taking a look.

from sysinfo.

Just to be sure: can you run process-viewer and htop alongside with the program highlighted to have a comparison please?

(process-viewer is using sysinfo so it'll make the bug more obvious and maybe easier for me to track)

from sysinfo.

I've used the example crate which seems to be working correctly and narrowed it down to total_time not being measured correctly. Just dumping values in terminal:

            let (new, old) = get_raw_times(&self.processors[0]);
            println!("New {:?} Old {:?}", new, old);

I get:

New 48856671 Old 48849869
Total time 6802, Use time 4509

In the example crate, but in our project:

New 48842622 Old 0
Total time 48842624, Use time 4136

Digging deeper to figure out why is that.

from sysinfo.

Tested all scenarios:

• Calling refresh_all before each refresh_process produces correct results edit: actually incorrect.
• Calling refresh_all once on init makes the total_time stuck on the same value for all refresh_process calls.
• Calling refresh_all and then probing the process without calling refresh_process produces values that are too large edit: actually correct.

from sysinfo.

Those are quite weird results... But it means that refresh_all makes an update needed for refresh_process. It reduces the possibilities quite a lot. With this, I think I'll be able to fix the issue pretty quickly. Thanks a lot!

from sysinfo.

Cheers, thanks for doing this!

edit FYI: I can reproduce all of that behavior in the example crate.

from sysinfo.