Why get processName from /proc/(pid)/comm should be at most 14 characters ?

[JimChenWYU]

Describe the bug
Why get processName from /proc/(pid)/comm should be at most 14 letters ?

Refs https://github.com/shirou/gopsutil/blob/master/process/process_linux.go#L795

To Reproduce

1. My python script named admin_worker.py, just example

#!/usr/local/easyops/python/bin/python
# -*- coding: utf-8 -*-

from time import sleep

while True:
  print("process...")
  sleep(1)

2. Start admin_worker.py

$ chmod +x admin_worker.py
$ ./admin_worker.py

3. In Linux, running my go program

// `pid` should be found by `ps` command in machine.

proce, _ := process.NewProcess(int32(pid))
name, _ := proce.Name()
fmt.Println(name)

I find the name is /usr/local/easyops/python/bin/python

Expected behavior
I expect the name is admin_worker.py, because i use cat /proc/(pid)/comm in machine, i get my expected value.

Environment (please complete the following information):

• Windows: [paste the result of ver]
• Linux: [paste contents of /etc/os-release and the result of uname -a]
• Mac OS: [paste the result of sw_vers and uname -a
• FreeBSD: [paste the result of freebsd-version -k -r -u and uname -a]
• OpenBSD: [paste the result of uname -a]

os-release

uname -a

Additional context
[Cross-compiling? Paste the command you are using to cross-compile and the result of the corresponding go env]

No.

[shirou]

Umm. I faintly think there was a reason, but I can't recall the details, and I can not find any comments from GitHub. If no reason, we can remove the limitation.

[Lomanic]

An easy reproducer for python psutil

TMPDIR=$(mktemp -d)
python3 -m venv "$TMPDIR/venv"
"$TMPDIR/venv/bin/pip3" install psutil
cat > "$TMPDIR/loooooooooooooooooooong.py" <<EOF
#!$TMPDIR/venv/bin/python3
import psutil
print(psutil.Process().name())
EOF
chmod +x "$TMPDIR/loooooooooooooooooooong.py"
cp -a "$TMPDIR/loooooooooooooooooooong.py" "$TMPDIR/short.py"
"$TMPDIR/loooooooooooooooooooong.py"
# returns loooooooooooooo
"$TMPDIR/short.py"
# returns short.py

psutil uses /proc/$PID/stat on linux, see https://github.com/giampaolo/psutil/blob/962cb9ad58ca3bc21b20130f435f8d507d931051/psutil/_pslinux.py#L1745 and https://github.com/giampaolo/psutil/blob/962cb9ad58ca3bc21b20130f435f8d507d931051/psutil/_pslinux.py#L1686. man proc documents this property in the /proc/$PID/stat file with

              (2) comm  %s
                     The  filename  of  the executable, in parentheses.  Strings longer than TASK_COMM_LEN (16) characters (including the terminating null byte) are silently truncated.
                     This is visible whether or not the executable is swapped out.

I can't test psutil behaviour on bsd (including darwin), where gopsutil behaviour (aligned with linux, using cmdline as a fallback) is also probably similarly broken (falling back to basename(cmdline[0]), see #796/#795.

What should be done? This is misaligned with psutil (and inconsistent) but changing this behaviour in gopsutil to be the same as in psutil will break backward compatibility downstream.

[JimChenWYU]

Refs

• https://gitlab.com/procps-ng/procps/-/commit/3e1c00d051ae4a8319df7714d0545b4b7aa81557
• https://gitlab.com/procps-ng/procps/-/commit/bb272580a6f3c192c61e307f9e341514d892d06e