| name | linux-tmux-dracula-cpu-temp |
| description | Show CPU temperature in the tmux status bar via the Dracula theme, with dynamic color based on the value, on Linux (desktops, servers, SBCs). Update-safe — uses /sys/class/thermal/ sysfs and does not patch the dracula plugin. Use when working with tmux on any Linux box and the user mentions adding a CPU temp segment to the tmux status line. |
Show Linux CPU temp in the tmux status bar (Dracula, colored, update-safe)
Companion to the macOS version (tmux-dracula-cpu-temp). The
mechanism is the same — append a status-right segment after the
tpm run line so dracula stays untouched — but the temp source is
the kernel sysfs thermal zone, which exists on most Linux systems
(desktops, servers, and SBCs alike). KVM guests usually don't
expose it.
Tested on:
- Raspberry Pi Zero W (armv6, BCM2835)
- Raspberry Pi Zero 2 W (aarch64, BCM2710A1)
- Orange Pi Zero 2 W (aarch64, Allwinner H618)
1. Find the right thermal zone
The mapping of thermal_zone0, thermal_zone1, etc. is not stable
or standardized across systems. On one machine thermal_zone0 might
be the CPU package; on another it could be the ACPI chassis sensor,
the WiFi card, or a battery sensor. Don't assume thermal_zone0 is
always the CPU.
To find what each zone represents, check its type:
for z in /sys/class/thermal/thermal_zone*; do
echo "$z: $(cat $z/type) = $(cat $z/temp)"
done
The temperature value is in millidegrees Celsius (e.g. 45000 =
45.0°C).
Look for a type like x86_pkg_temp, cpu-thermal, acpitz, or
similar — that's the one to use. The script below defaults to
thermal_zone0; adjust the path if your CPU is on a different zone.
If no thermal zone exists at all, the box doesn't expose CPU
temperature (typical for cloud KVM guests like Oracle Cloud,
RackNerd, etc.). On RPi you can also use vcgencmd measure_temp,
but the sysfs path is universal.
2. Wrapper script
~/.tmux/scripts/cpu_temp.sh (POSIX sh, no bashisms — works on
busybox / dash too):
#!/bin/sh
f=/sys/class/thermal/thermal_zone0/temp
if [ ! -r "$f" ]; then
printf '#[fg=#282a36,bg=#6272a4] |n/a'
exit 0
fi
t=$(awk '{printf "%.0f", $1/1000}' "$f")
if [ "$t" -lt 50 ]; then bg='#8be9fd'; ico='🥶'
elif [ "$t" -lt 65 ]; then bg='#50fa7b'; ico='😎'
elif [ "$t" -lt 75 ]; then bg='#ffb86c'; ico='🥵'
else bg='#ff5555'; ico='🔥'
fi
printf '#[fg=#282a36,bg=%s] |%s%s°C' "$bg" "$ico" "$t"
chmod +x ~/.tmux/scripts/cpu_temp.sh
Bands chosen for passively-cooled SBCs:
<50°C cold (cyan)50–64°C normal (green)65–74°C warm (orange)≥75°C hot (red) — RPi soft-throttles at 80°C, so 75 is a sane warning band.
3. Wire it into ~/.tmux.conf
At the very bottom, after the tpm run line:
run '~/.tmux/plugins/tpm/tpm'
# Append CPU temp segment to status-right (after dracula has set it).
# Script emits its own color based on temp value. Survives TPM updates.
set -ag status-right "#(~/.tmux/scripts/cpu_temp.sh)"
4. Reload (and clean up duplicates if needed)
tmux source-file ~/.tmux.conf
Gotcha: duplicated segments after deploy
If b0o/tmux-autoreload is enabled (very common in dracula setups)
it will detect the conf edit and re-source in parallel with your
manual source-file. Both runs race past dracula's
set -g status-right "" reset, leaving you with 2× or 3× of each
segment. Fix:
tmux set -g status-right ""
tmux source-file ~/.tmux.conf
Verify only one of each appears:
tmux show -gv status-right | grep -oE 'cpu_info|ram_info|cpu_temp' | sort | uniq -c
Gotcha: emoji width on narrow terminals (phone SSH)
This one bites hard. On a ~40-column phone SSH session the status line
starts wrapping its last 1-2 characters onto a second row, and every
5-second redraw adds another wrapped row until the screen is a stack
of half-drawn status bars.
Root cause: tmux and your terminal disagree on how wide an emoji is.
tmux measures the status with its own Unicode width table; some emoji it
counts as 1 cell while your terminal renders them as 2. tmux
thinks the line fits, doesn't truncate, and the terminal then paints it
1 cell too wide -- and because the status row sits at the bottom with
autowrap on, it wraps and scroll-accumulates on every refresh.
It is not all emoji -- only the ones with Unicode
Emoji_Presentation=No (legacy "text-presentation" pictographs):
🌡 ☀ ❄ ♨ ✈ ☁ ❤ ✏. tmux follows the spec and
counts these 1 cell; terminals render them as 2. Emoji with
Emoji_Presentation=Yes (🔥 🥵 🥶 😎 💻 🧊 …) are
counted 2 by tmux, matching the terminal -- those are safe.
The thermometer 🌡 (U+1F321) is one of the bad ones, which is why an
earlier version of this skill overflowed. The fix: use an
Emoji_Presentation=Yes glyph instead. The script above picks a
width-2 face per temperature band (🥶/😎/🥵/🔥), so tmux and the
terminal always agree and the status never overflows.
Things that do NOT work
- Variation Selector-16 (U+FE0F): appending
️ to force emoji
presentation does not change tmux's width count (verified on tmux
3.6b -- 🌡️ still measures width 1).
- Trimming spaces: shaving a space compensates for one bad emoji
but is fragile -- add another text-presentation emoji and it breaks
again. Fix the glyph, not the spacing.
Probe any glyph's tmux width
Stash the string in a user option and pad it with #{p<N>:...}, which
uses tmux's own width math (a bare literal after p: is treated as a
variable name, so the option indirection is required):
tmux set -g @m '🌡'
tmux display -p '[#{p8:#{@m}}]'
tmux set -gu @m
Width 1 = will overflow on a 2-cell-rendering terminal; width 2 = safe.
The general cure (terminal side)
The disagreement is really the terminal violating Unicode's
default-presentation rule. Many terminals (Blink, Termius, WezTerm,
kitty) expose a Unicode-width / "ambiguous = narrow" setting. Making the
terminal spec-compliant renders text-presentation emoji as 1 cell too --
then tmux and terminal agree for every emoji and you can use any
glyph, thermometer included.
Related
See tmux-dracula-cpu-temp for the macOS / Apple Silicon variant
(uses smctemp instead of sysfs, with higher temp bands).
