top: implement SUMMARY display

[Bluemangoo]

Resolves #207

Tasks:

• time and uptime
• active user count
• cpu load average
• task summary
• cpu load
• memory and swap summary
• determine memory unit from --scale-summary-mem

I add crate systemstat for some of the values.

top - 13:03:00 up 1:33, 1 user, load average: 2.01, 1.48, 1.28
Tasks: 1040 total, 6 running, 974 sleeping, 0 stopped, 0 zombie
%Cpu(s):  13.9 us, 3.3 sy, 0.0 ni, 82.4 id, 0.3 wa, 0.0 hi, 0.1 si, 0.0 st
MiB Mem :   7636.4 total,    433.3 free,   6415.1 used,    788.0 buff/cache
MiB Swap:  30518.0 total,  27218.5 free,   3299.5 used,   1221.3 avail Mem

[Krysztal112233]

That's great. But can I ask if it can you use text instead of image? That would be more search friendly.

[Bluemangoo]

There are no way to measure load average on Windows. We can only get an instantaneous CPU load with this command:

wmic cpu get loadpercentage

The Python wheel psutil emulates what Linux does (for sure, need time to collect data):

https://github.com/giampaolo/psutil/blob/d4b37d9628e634c00a2990e488b1cc38ea6a41b5/psutil/arch/windows/wmi.c#L16-L54

I'm not sure what should we do: ignore it, or do what psutil does.

tips: psutil doesn't work well on my computer. it shows (0.0, 0.0, 0.0)

[Krysztal112233]

In fact, the crate we used required us does the same thing:

https://github.com/uutils/procps/blob/c2c0baba2d0db19f40c04f86df073c5614333858/src/uu/top/src/top.rs#L53C1-L57C54

    // Must refresh twice.
    // https://docs.rs/sysinfo/0.31.2/sysinfo/struct.System.html#method.refresh_cpu_usage
    picker::sysinfo().write().unwrap().refresh_all();
    sleep(Duration::from_millis(200));
    picker::sysinfo().write().unwrap().refresh_all();

You can even use a multi-threaded background to refresh this data if sysinfo cannot doesn't meet your needs.

[Bluemangoo]

i add libsystemd-sys to get correct active users count(on linux).

starting with systemd 256.7-2, utmp support has been disabled. This
was done as utmp is not Y2038 safe.

After this change, /run/utmp is no longer created by systemd

i check if system is booted by systemd, if so get active user count from it, or from utmp, like what gnu does.

so the apt package libsystemd-dev is required. shall i install it in workflows?

[Krysztal112233]

i add libsystemd-sys to get correct active users count(on linux).

starting with systemd 256.7-2, utmp support has been disabled. This
was done as utmp is not Y2038 safe.

After this change, /run/utmp is no longer created by systemd

i check if system is booted by systemd, if so get active user count from it, or from utmp, like what gnu does.

so the apt package libsystemd-dev is required. shall i install it in workflows?

Is there any other way to detect the system is initialized with systemd? (e.g. some features on the Init process?)

[Krysztal112233]

I think using libsystemd-sys may introduce additional compilation dependencies that aren't necessary...and identify them by identifying some features might be better such as

➜  ~ cat /proc/1/status  
Name:	systemd
Umask:	0000
State:	S (sleeping)
Tgid:	1
Ngid:	0
Pid:	1
PPid:	0
TracerPid:	0
Uid:	0	0	0	0
Gid:	0	0	0	0
FDSize:	256
Groups:	 
NStgid:	1

As you can see the field Name is systemd.

This method cannot work on other system so you have to use something such as #[cfg] to support other system, support Linux is enough currently.

[Bluemangoo]

Yes i used #[cfg].

https://github.com/Bluemangoo/procps/blob/9b45c481c1450371f2c9f3c54ab4e64ad290e335/src/uu/top/src/header.rs#L157

so as the dependency.

https://github.com/Bluemangoo/procps/blob/9b45c481c1450371f2c9f3c54ab4e64ad290e335/src/uu/top/Cargo.toml#L25

if system is booted by systemd, sd_booted will return a number bigger than 0.

https://github.com/Bluemangoo/procps/blob/9b45c481c1450371f2c9f3c54ab4e64ad290e335/src/uu/top/src/header.rs#L165

[Bluemangoo]

I would like to mention that platform-specific dependencies do not work with workspace dependencies. Therefore, I specify the version of these dependencies in top's Cargo.toml file, rather than that in the root.

[Bluemangoo]

The way CPU time is calculated in Windows differs from that in Linux. On Windows, CPU time is divided into the following categories: idle, kernel, user, dpc, and interrupt. The dpc time is similar to Linux's soft interrupt, while the interrupt time is specifically for hardware.

Should I display them in the following format:
%Cpu(s): 0.7 us, 50.0 sy, 49.3 id, 0.0 hi, 0.0 si
or in this format:
%Cpu(s): 0.7 us, 50.0 sy, 49.3 id, 0.0 hi, 0.0 dpc?

Additionally, in the Windows documentation, it mentions reserved fields (dpc, interrupt) in the struct:

typedef struct
_SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION {
    LARGE_INTEGER IdleTime;
    LARGE_INTEGER KernelTime;
    LARGE_INTEGER UserTime;
    LARGE_INTEGER Reserved1[2];
    ULONG Reserved2;
} SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION;

I found a possible explanation in psutil.

typedef struct {
    LARGE_INTEGER IdleTime;
    LARGE_INTEGER KernelTime;
    LARGE_INTEGER UserTime;
    LARGE_INTEGER DpcTime;
    LARGE_INTEGER InterruptTime;
    ULONG InterruptCount;
} _SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION;

[sylvestre]

please let me know when I can review it. It doesn't have to be completed to land!

[Bluemangoo]

It seems that the load average on Windows may require opening a new issue. The CPU load for macOS is still to be done.

you can begin reviewing the rest of the content.

[Bluemangoo]

all fixed.

[Krysztal112233]

Here's some bug comes from upstream, this pr will be merged after new version of uucore released.

[Bluemangoo]

not severe issue, you can also merge first, and update later. I hope #292 is not waiting for this PR.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 0.00%. Comparing base (16e31d6) to head (db8ecf5).
Report is 26 commits behind head on main.

Additional details and impacted files

@@     Coverage Diff     @@
##   main   #306   +/-   ##
===========================
===========================

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.