[Impact]
A memory leak in tmux causes it to eventually fail when run for long periods of time. Since long term tmux sessions are common, this becomes a real problem for sysadmins, etc.
[Test Case]
# apt update
# apt dist-upgrade -y
# apt install -y tmux
# tmux -L none -f /dev/null new-session -s leaky-tmux-test 'tmux set-option -w -g pane-border-status top; tmux pipe-pane "cat>/tmp/leaky-tmux-test.log"; p=${TMUX#*,}; p=${p%,*}; s=0; j=0; echo "ITER: RSZ"; while [ $j -lt 6000 ] ; do j=$((j+1)); n=$(ps hwo rsz $p); printf "\e]2;%d: %d\e\\\\\r%4d: %d" $j $n $j $n; [ $s -ne $n ] && { echo; s=$n; }; done'
[exited]
# cat /tmp/leaky-tmux-test.log ; echo
ITER: RSZ
1: 3712
1360: 3776
2725: 3904
4215: 4032
5743: 4160
6000: 4160 (example output)
RSZ will only change during the test if there's a memory leak. By comparison, we get correct behavior running without pane-border-status top:
$ tmux -L none -f /dev/null new-session -s leaky-tmux-test 'tmux pipe-pane "cat>/tmp/leaky-tmux-test.log"; p=${TMUX#*,}; p=${p%,*}; s=0; j=0; echo "ITER: RSZ"; while [ $j -lt 6000 ] ; do j=$((j+1)); n=$(ps hwo rsz $p); printf "\e]2;%d: %d\e\\\\\r%4d: %d" $j $n $j $n; [ $s -ne $n ] && { echo; s=$n; }; done'
[exited]
$ cat /tmp/leaky-tmux-test.log ; echo
ITER: RSZ
1: 3492
6000: 3492
[Regression Potential]
Since this is only plugging a memory leak, no other behavior should be affected.
[Original Description]
The upstream bug is reported and fixed at https://github.com/tmux/tmux/issues/1167.
Can we either patch the package (patch at https://github.com/tmux/tmux/commit/c363c236aaea5b7a879493d8f3c85bead546f063) or upgrade to 2.7 (I guess this is less likely to happen given the imminent release).
Hi antgel,
I'm passing a few bugs with not enough attention atm.
Your report is great - I put it in a queue to hopefully be picked up soon.
Do you have a testcase other than waiting to see the issue?
It is in screen_
I tried
$ while /bin/true; do tmux refresh-client; done
but that did not expose the issue.