#procps 4.0.6 is now out along with #debian packages

What's new?
Should work better with elogind
w -t terminal mode for terminals
w "finding the tty" improved
watch 256 color support
w -pt doesn't crash it

https://gitlab.com/procps-ng/procps/-/releases/v4.0.6
#opensource #linux

Another interesting procps bug. If you try to match a process by tty, you may get more than you expect.

This is because the match is done by device ID. If you have a mount namespace, this can mean you have more than one, say /dev/pts/0 so if you run pgrep on that tty, it might show processes within containers.

One fix is to check you are in the same mount namespace, but means if you do start a docker process and run ps it won;t show up.
#procps #programming

Today I learnt about stable and unstable sorting algorithms. Stable means for items with the same key, the order those items came into the sort remain the same.

In ps if you ask to sort by two keys, this matters. If you ask to sort by user and start time, the start time sort is done first, then user. But the user sort can scramble the start time (within a user's processes) because qsort() is not stable.

#procps #linux #programming

Finally got to the bottom of a ps bug. If you use some, but not all, forest modes then direct children of init will not show up foresty.
It seems there was a check in the forest code where if the pid was 1 and forest mode was not unix, then the children didn't get the initial "tree branch" ascii art.

I tried tracking back to see why, got to 2002 and it was still there so gave up and fixed it.

#opensource #procps