N?? / show marks the upstream candidate with "*" instead of ">" (the
prompt, which read as a command in logs), and every line but the note
gets a two-character gutter: " " for the installed record, "* " for the
candidate, so keys and colons line up. The plain native fallback (zypper,
unmergeable records) and the not-installed candidate use the same gutter.
pacman records (-Qi with -Si or the AUR helper's -Sia) are now compared
the way they are meant:
- values unwrapped (continuation lines joined, runs of blanks as one):
yay wraps Depends On / Optional Deps at 80 columns even without a
terminal unless COLUMNS=0, which the helper now gets for -Qua and -Sia
- Depends On, Provides, Conflicts With, Make/Check Deps and the other
lists as sets of exact tokens in any order (a constraint is a change)
- Optional Deps by package names ("name: description [installed]" in
-Qi, bare names from the AUR)
- a differing value is one line like -Qi, Optional Deps one entry per
line; candidate-only fields sit next to their counterpart (Repository
after Version, AUR URL after URL, Maintainer after Packager, ...)
Verified on recorded claude-desktop records (paru 2.1.0, yay 13.0.1, the
installed 2.9939.4-1 from the AUR's own .SRCINFO); discovery/pacman.md
section 15.2.
On the pacman backend, an installed AUR helper (paru, else yay;
--aur-helper=paru|yay|none, --no-aur) adds the AUR updates to the review
list. The helper never runs as root: when the tool is root through sudo it
runs as SUDO_USER (runuser -u, else sudo -u, with that user's HOME, from
that home); unprivileged it runs as is; a root login gets no AUR, with a
warning. The banner names the helper ("+ AUR via paru").
- listing: <helper> -Qua (also with --no-refresh), rows with repository
"aur"; [ignored] is shown [held]; a failing -Qua warns and the session
goes on with the repository rows
- every row has a marker column after the number: "*" for AUR rows, a
space otherwise (on every backend)
- N?? merges pacman -Qi with <helper> -Sia; N?, N+, Nx stay with pacman;
N- excludes through the helper's --ignore
- <enter>: the repository step exactly as before, then <helper> -Sua
--ignore <excluded> as the invoking user (--noconfirm only when stdin is
not a terminal); pacman is skipped when only AUR rows are left; a helper
failure is reported and its status returned; --dry-run prints it
- other backends never run a helper; --aur-helper is ignored with a warning
paru 2.1.0 and yay 13.0.1 formats verified in discovery/pacman.md section
15 (proof: discovery/pacman/aur.sh). Unit tests replay the recorded output
(fake paru, yay, runuser, getent); the archlinux E2E run installs the
foreign su-dummy-aur and drives a fake paru through a fake sudo built on
passwordless su for wheel.
N?? and `show PKG` now print the installed record in the package
manager's own key/value layout, after one line explaining the format:
# lines starting with > come from the upstream candidate: a differing
value, or a field the installed record lacks
After each field whose candidate value differs comes the candidate's
line prefixed with "> "; fields only the candidate has are appended the
same way; installed-only fields stay; identical fields print once.
Multi-line values (pacman Optional Deps, dnf Description, RFC-822
continuation lines) are compared as blocks and a differing candidate
block gets "> " on every line. A package that is not installed shows the
candidate alone ("# not installed: this is the upstream candidate").
Merged where the backend gives comparable records for both sides:
- pacman: -Qi vs -Si REPO/PKG, REPO = the first repository -Sl lists
- apt: dpkg-query -s (Status, Conffiles, the long description; richer
than apt-cache show PKG=<installed>, which gives the archive record
when that version is still listed) vs apt-cache show --no-all-versions;
Description-xx pairs with Description and a one-line candidate
Description equal to the installed first line counts as identical
(the archives ship the long text in Translation files)
- dnf: info --installed vs info --upgrades (--available lists other
architectures first and nothing for an up-to-date package); not
installed: --available; dnf5's differently padded keys are re-padded
zypper keeps `zypper info --requires`: rpm -qi and zypper info share
almost no keys and zypper cannot print the installed record itself.
No candidate, or not exactly one record per side, keeps the old output.
Formats verified in containers and recorded in discovery/*.md. Unit
tests replay recorded pacman, apt, dnf4 and dnf5 records; the E2E suite
checks 2?? on su-dummy-b on all 12 distros.
When stdin is a terminal the prompt is read with bash's readline,
`IFS= read -e -r -p '> ' REPLY`, so the arrow keys no longer print escape
garbage: Left/Right, Home/End, Ctrl-A/E, Backspace/Delete edit the line,
and Up/Down walk the lines entered in this session (`history -s` for every
line with a non-blank character; empty lines are not recorded). Down past
the newest entry restores the line typed before Up, and Up at the oldest
entry or Down on the new line ring the bell, as readline does by default.
The Nx y/N answer is read the same way and never recorded.
Nothing persists: HISTFILE is unset at startup and the history is never
read or written. Verified on bash 4.3.48 and 5.2.15 that `read -e` and
`history -s` share one history list in a non-interactive shell, that
inherited HISTSIZE/HISTCONTROL/HISTIGNORE have no effect there, and that
Ctrl-C at the prompt restores the terminal. readline draws the prompt on
stderr, as bash does. Piped stdin keeps the plain `read -r` exactly,
including the echo of each line and EOF => exit 2.
New unit tests (tests/unit/lineedit.bats) drive a real pty with script(1):
history recall, the bell at both ends, the stashed line, Left-arrow
editing, the y/N answer kept out of the history, no history file under
HOME/TMPDIR, ^D => exit 2. The E2E suite gets a pty smoke test (Up
recalls 1?), skipped where the image has no script(1).
N? and `info PKG` now print three lines, none indented:
PKG: <description>
reason: explicitly installed | installed as a dependency | unknown | not installed
required by: <installed dependents, comma separated> | nothing
The reason comes from a new <be>_reason backend function: apt-mark
showauto / showmanual (dpkg decides for a package in neither list), the
Install Reason of pacman -Qi, dnf repoquery --installed %{reason} (dnf4
and dnf5 words) and the Installed line of zypper info (rpm plus libzypp's
AutoInstalled when zypper is locked). A row whose old version is "-" is
not installed and the backend is not asked. The "(installed)" qualifier
moves to --help, README and DESIGN.
The rpm E2E fixture and e2e_reset now also mark su-dummy-a and su-dummy-c
as user-installed on dnf, which records plain rpm installs as "unknown",
so 1? and 2? show both reasons on all 12 distros. The verified commands
are recorded in discovery/{apt,pacman,rpm}.md.
bryanpedini