Incomprehensible issue after adding a Lua script using lua_load

[Skif-off]

My old config:

conky.config = {
	background = true,
	double_buffer = true,

	alignment = 'top_right',

	border_width = 1,
	cpu_avg_samples = 2,
	default_color = 'white',
	default_outline_color = 'white',
	default_shade_color = 'white',
	draw_borders = false,
	draw_graph_borders = true,
	draw_outline = false,
	draw_shades = false,

	gap_x = 10,
	gap_y = 90,
	net_avg_samples = 2,
	no_buffers = true,
	out_to_console = false,
	out_to_stderr = false,
	extra_newline = false,

	own_window = true,
	own_window_type = 'normal',
	own_window_transparent = true,
	own_window_argb_visual = true,
	own_window_argb_value = 0,
	own_window_hints = 'undecorated,below,sticky,skip_taskbar,skip_pager',

	minimum_width = 210, minimum_height = 0,
	stippled_borders = 0,
	update_interval = 2.0,
	uppercase = false,
	use_spacer = 'none',

	show_graph_scale = false,
	show_graph_range = false,

	use_xft = true,
	text_alpha = 0.1,
	font = 'Liberation Mono:size=10',
	color0 = 'white',
	color1 = '#e69300'
};

conky.text = [[
${color1}CPU ${color0}${alignr}${cpu cpu0}%
${cpubar cpu0}
${top name 1} ${alignr}${top cpu 1}%
${top name 2} ${alignr}${top cpu 2}%
${top name 3} ${alignr}${top cpu 3}%
${top name 4} ${alignr}${top cpu 4}%
${top name 5} ${alignr}${top cpu 5}%

${color1}RAM ${color0}(${memmax}) ${color0}${alignr}${memperc}%
${membar}
${top_mem name 1} ${alignr}${top_mem mem_res 1}
${top_mem name 2} ${alignr}${top_mem mem_res 2}
${top_mem name 3} ${alignr}${top_mem mem_res 3}
${top_mem name 4} ${alignr}${top_mem mem_res 4}
${top_mem name 5} ${alignr}${top_mem mem_res 5}

${color1}RAM use/free:
${color0}${alignr}${mem}/${memfree}
${color1}Classic:
${color0}${alignr}${legacymem}/${memeasyfree}

${color1}SWAP ${color0}(${swapmax})${color0}${alignr}${swapperc}%
${swapbar}
 ${color1}Use/Free: ${color0}${swap}/${swapfree}

${color1}Processes: ${color0}${processes} ${color1}Running: ${color0}${running_processes}
${color1}________________________
${color1}CPU Frequency:
${color0}CPU1: ${alignr}${cpu cpu1}% ${freq 1} MHz
${color0}CPU2: ${alignr}${cpu cpu2}% ${freq 2} MHz
${color0}CPU3: ${alignr}${cpu cpu3}% ${freq 3} MHz
${color0}CPU4: ${alignr}${cpu cpu4}% ${freq 4} MHz
${color0}CPU5: ${alignr}${cpu cpu5}% ${freq 5} MHz
${color0}CPU6: ${alignr}${cpu cpu6}% ${freq 6} MHz
${color0}CPU7: ${alignr}${cpu cpu7}% ${freq 7} MHz
${color0}CPU8: ${alignr}${cpu cpu8}% ${freq 8} MHz

${color0}${exec lua $HOME/.config/conky/run_sensors.lua}
${color1}________________________
${color1}Kernel: ${color0}${kernel}
${color1}Uptime: ${color0}${uptime_short}
]];

run_sensors.lua:

-- run_sensors.lua
-- 2026.10.03
--[[
For Conky: getting some info from "sensors" output.

Motherboard: Asus M5A97 R2.0
CPU: AMD FX-8320

Don't forget to load kernel module for ISA adapter (it8721)
  it87: IT8705F/IT871xF/IT872xF hardware monitoring driver
]]

local h = io.popen("sensors")
local out = h:read("*a")
h:close()
local n = 1
local cf, cw, t1, t2, ct, gt, gf

n = string.find(out, "radeon-pci", 1, true)
gt = string.match(out, "temp1: +(%+%d+%.%d+°C)", n)
gf = string.match(out, "freq1: +(%d+ MHz)", n)

n = string.find(out, "fam15h_power-pci", 1, true)
cw = string.match(out, "power1: +(%d+%.%d+)", n)

n = string.find(out, "it8721-isa", 1, true)
if n == nil then
  cf, t1, t2 = "", "", ""
else
  cf = string.match(out, "fan1: +(%d+ RPM)", n)
  t1 = string.match(out, "temp1: +(%+%d+%.%d+°C)", n)
  t2 = string.match(out, "temp2: +(%+%d+%.%d+°C)", n)
  if t2 == nil then t2 = "ERROR" end
end

n = string.find(out, "k10temp-pci", 1, true)
ct = string.match(out, "temp1: +(%+%d+%.%d+°C)", n)

print("Power: 125 / " .. cw .. " W")
print("CPU temp: " .. ct)
print("it8721:")
print("   temp1: " .. t1)
print("   temp2: " .. t2)
print("CPU fan: " .. cf)
print("GPU freq: " .. gf)
print("GPU temp: " .. gt)

I decided to get rid of the ${exec} variable, I found a couple of Lua modules for working with libsensors, but I thought: /sys resides in RAM, so why not simply read the required values?
I added lua_load = '4hwmon.lua'to the conky.config section and replaced ${exec} with

${color0}Power: ${alignr}125 / ${lua 4hwmon fam15h_power power 1} W
${color0}CPU temp: ${alignr}+${lua 4hwmon k10temp temp 1}°C
${color0}it8721, temp1: ${alignr}+${lua 4hwmon it8721 temp 1}°C
${color0}it8721, temp2: ${alignr}+${lua 4hwmon it8721 temp 2}°C
${color0}it8721, Vcore: ${alignr}${lua 4hwmon it8721 in 2} V
${color0}CPU fan: ${alignr}${lua 4hwmon it8721 fan 1} RPM
${color0}GPU
  temp: ${alignr}+${lua 4hwmon radeon temp 1}°C
  freq: ${alignr}${lua 4hwmon radeon freq 1} MHz

and 4hwmon.lua:

-- 2026.10.06
--[[
Idea based on https://github.com/shikhalev/conky/blob/main/lua/hwmon.lua
]]

local hwmon_indices = {}
local idx = 0
while true do
  local p = "/sys/class/hwmon/hwmon" .. idx
  local h = io.open(p .. "/name", "rb")
  if h == nil then break end
  local s = h:read("*a")
  h:close()
  local n = string.find(s, "\n", 1, true)
  s = string.sub(s, 1, n - 1)
  hwmon_indices[s] = idx
  idx = idx + 1
end

function GetData(fn)
  local h = io.open(fn, "rb")
  if h == nil then return nil end
  local s = h:read("*a")
  local n = string.find(s, "\n", 1, true)
  if n == nil then return nil end
  return tonumber(string.sub(s, 1, n - 1))
end

function conky_4hwmon(device, tp, index)
  local r = GetData("/sys/class/hwmon/hwmon" .. hwmon_indices[device] .. "/" .. tp .. index .. "_input")
  if r == nil then return "FAIL!" end
  if tp == "power" then
    return string.format("%.2f", r * 0.000001)
  elseif tp == "in" then
    return string.format("%.2f", r * 0.001)
  elseif tp == "freq" then
    return string.format("%d", r * 0.000001)
  elseif tp == "temp" then
    return string.format("%.1f", r * 0.001)
  else
    return r
  end
end

And ended up with a problem: at some point, the data of all Conky variables disappears and they start displaying again after a few cycles (usually between 2 and 15 seconds, with update_interval = 2.0 it will take 2–7 cycles). (Sometimes only part of the data is displayed initially, but the full set appears on the next cycle.)
However, after auto-restoring, the ​​${freq n} values are not displayed, so instead of CPUXn: 100% 3500 MHz Conky shows CPUXn: 100% MHz. The only solution is restart Conky (or PC :))).

Consistently reproducible in 1.25.3 and 1.24.2; less frequently in 1.23.0.
Has not yet been reproduced in 1.22.3.
I tried conky --debug, but didn't see anything useful.

Usually, some kind of load is required. For example: run VirtualBox (I use VM to build packages/applications) > Update the list of available packages and start the installation of the specified ones > the packages are downloaded, and their unpacking begins. > problem usually arises at this point. Or I can run multiple copies of 7z b.
It is easier to reproduce with version 1.25.3; a more modest load is sufficient, for example, 20 minutes ago, on a freshly booted system, I simply launched Double Commander, opened a Lua script in Geany, started Firefox ESR, and opened the https://lua.org URL...

Any ideas?

P.S. Xubuntu 24.04, X11. If possible, I prefer packages; build options can be viewed here (see the build-all task).

[Skif-off]

1.22.3 with the ame behavior.

Is it looks like I did something wrong, not Conky? But I don't see yet where I went wrong...

[Skif-off]

I caught it, with conky --debug:

[error] [src/common.cc:235] can't open file '/proc/uptime': Too many open files
[error] [src/common.cc:235] can't open file '/proc/meminfo': Too many open files
[error] [src/common.cc:235] can't open file '/proc/meminfo': Too many open files
[error] [src/common.cc:235] can't open file '/proc/stat': Too many open files
[error] [src/common.cc:235] can't open file '/proc/cpuinfo': Too many open files
[error] [src/data/os/linux.cc:1618] failed to access '/proc/cpuinfo' at get_freq: Too many open files

The default limit for nofile is 1024 (soft).

I’ll check how files in /sys/class/hwmon/hwmonN are updated and whether it’s possible to simply re-read them without closing the file descriptors (i.e. file:seek("set") and file:read() again, without file:close()).

[Skif-off]

About descriptors :)) I noticed completely by chance, using peripheral vision, that the GetData function looks shorter than it should be (in terms of the number of actions).
It seems that this is the first time I’ve forgotten to add h:close or [more likely] made a mistake when I corrected the script after testing (I was checking whether string.find + string.sub would work faster than string.match).
I’ll keep an eye on it, and if everything is fine, I’ll close the discussion.

[Skif-off]

Thanks for PR#2493.