This adds a MvRn linearly weighted moving average of PRn (processes marked runnable) for a more perfectly precise load average than the usual 1 min/5min/15 min EWMA load averages of the 1970s.

This LWMA goes to exactly zero after the (usually small-ish) measurement window is expired. Default data window is the same as frqHdr so from the same row in one bank of lines to the one "1 frame ahead" is the measurement window which is a nice visual cue, although it is independently settable.

Also merge IrJ+nice+steal into a new OthJ other-jiffie variable so total non-idle CPU has a 4 column breakdown (used in the default format). Because the MvRn is usually all you want (and really the 4 jiffies are way more informative anyway), PRn itself will probably leave the default format once I have some reason to want another data column.

Runnable|stopped is propated to whole groups so that if some child of a big merge group is stopped and you have stopped set to blink then you will see "something is stopped in the group". Similar for runnable-ness.

Recent work seems to have settled down. So, I'm releasing this now rather than leave unreleased config & parc updates around for a long time.

RELEASE-NOTES.md has full details and git log max detail, but in brief:

• Add dp, kp, fp, wp & k short aliases
• Add ability to visually indicate sort order & direction via -g,--glyph
• Fixed for both old & new header size-induced space insertion in headers so that e.g. pd|awk probably works

RELEASE-NOTES.md has more details and the git log full detail.

There is also a new configs/kaminsky0/add-final-header.awk program that may have more general utility than just for procs (basically for any table with headers that may sometimes be overtall - see the discussion here for more details / background.

RELEASE-NOTES.md has more details and the git log full details, but briefly:

• new program (both C & Nim version) util/parc allows fast, standalone cpio archive creation of /proc files that uses $j (as in j=-1 parc ...) to use multiple workers like make -j.
• procs find gets pidfd_send_signal, --if-handled, --run-state
• procs display gets --eqLeaf which makes sorting more useful combined with forest mode. This being true may become the default in the near future and also %D format converts to "" if order does not also start with D.
• pd also sorts before merging to get more deterministic roll-ups
• configs/cb0 has been updated quite a bit if you base your configs off of that

As usual, the release notes have more details and git log full details.

The title gives the headliner features and RELEASE-NOTES.md gives all the details.

Unobserved in git log or notes is the dramatic speed-up of a small mmap'd cpio archive compared to 10s of thousands of /proc fs calls which is, informally, over an order of magnitude.

RELEASE-NOTES.md has more details, but in brief:

• Add batch labeling to procs.find/pf via --Labels=
• Allow users to re-direct /proc inquiries to $PFS
• Make bench.sh more portable

There are still pending ideas for some kind of batch interface to using procs.find aka pf as a mechanism to type processes. This is a checkpoint release before those pf CL modifications.

This release:

• Makes labels emitted by pf and consumed by the main CL syntax :-delimited strings not chars
• In light of the above, the field width of the union of all labels is expanded to 30 for maybe 3..6 labels before messing up tabular alignment
• Uses a longer, but still bounded amount of %C for pcr_C
• Internally uses sysUt.!! to raise exceptions
• Updates a few things in the self-documentation
• Makes the bench.sh script more robust if bu/ tools are in $HOME/bin