Non-critical packages (Desktop/Plasma group, Utilities) went through try_install which installed them one-at-a-time (45 individual pacman calls). Override operation_try_install in PMPacman to batch all plain-string packages into a single call, matching the behavior of operation_install.
Also:
- Remove antergos-plasma-theme from rootfs (profile.yaml) and both netinstall Plasma groups
- Add allowFileSystemTypeSelection: true to both antergos and minimal partition.conf
- Both main.py files (profile + minimal) get the batching fix
so github told me my repo has NO community standards apparently
CODE_OF_CONDUCT? nonexistent. ISSUE_TEMPLATES? nada.
PULL_REQUEST_TEMPLATE? the void stared back.
and github profiles has a little checklist like im some
kind of degenerate who doesnt know how to run a project.
well NOW i have standards. and they are ROCK SOLID STABLE.
you want systemd? fork from before-systemd-change and figure
it out yourself, i gave you a numbered tutorial right in the
CoC. the AI moderator will read it and then politely but
firmly tell you to go fork yourself (the repo, not literally,
read the CoC).
also threw in the pixie SDDM switch that i apparently never
committed to the netinstall yamls either (kill me), the batch
install main.py cleanup, and fixed the README lying about
why artix.
DEUS VULT and please file your issues in the YAML forms
provided or the AI will judge you
Restores original callback (hardcoded False), pretty_status_message refactor,
umask pattern, known_inits name, unguarded operations access, and inline
subst_locale — all of which were changed in earlier CodeQL fix commits.
Keeps only harmless changes: unused import removal (sys, re), PEP8 comma
space, and _change_mode zero-division guard.
Adds large ASCII warning header: DO NOT EDIT without full ISO test.
- Guard _change_mode progress division against zero total_packages
- Document why -Syu is inappropriate in installer context (reproducibility)
- Promote known_inits to module-level KNOWN_INIT_PROVIDERS constant
All approaches to keep OpenRC as default failed:
- --ignore: pacman ignores it during fresh install provider resolution
- --assume-installed=init-logind: ALPM doesn't check assume-installed for virtual providers
- --ask=12: infinite loop (pacman re-adds elogind-dinit after each removal)
- dummy package in custom repo: init-rc conflict hits via different path
Dinit works because pacman --noconfirm naturally picks elogind-dinit
(alphabetically before elogind-openrc), pulling dinit-rc which has no
conflict. All inits (OpenRC, Runit, S6) remain selectable in the
online installer.
Changes:
- buildiso: INITSYS='dinit', remove --assume-installed/--ask hacks
- basestrap/main.py, packages/main.py: remove same hacks
- profile.yaml: antergos-live-dinit for dinit, move openrc pkg to init-specific
- packagechooser.conf (x2): default=Dinit, Dinit first in item list
- packagechooser_dm.conf: *-dinit instead of *-openrc
- README.md, README.pl.md, docs/*: update all OpenRC-first references
- AGENTS.md: document dinit as default
--ignore doesn't reliably prevent pacman from considering elogind-dinit
during fresh provider resolution on all pacman versions. --assume-installed
creates a dummy installed package for init-logind, satisfying the
dependency from base without needing any provider at all. The explicitly
listed elogind-openrc then gets installed normally and provides
init-logind for real.
Dropped --ignore entirely in favor of --assume-installed + --ask=4
(conflict resolution safety net). --assume-installed only added for
non-dinit providers (openrc, runit, s6).
The branding's netinstall.yaml has a Base System group with 'base'
that gets added to packageOperations. The packages module then
reinstalls base without --ignore, triggering the same init-logind
resolution bug. Adding the same --ignore logic from the basestrap
module.
Using --ignore=elogind-dinit (and alphabetically-before providers for
other inits) prevents pacman --noconfirm from resolving init-logind
to the wrong provider. This is more robust than installing the
provider separately, which can fail if base's deps are missing.
dinit-rc and openrc both provide AND conflict with virtual init-rc and
cannot coexist. When base depends on init-logind (virtual), pacman with
--noconfirm picks the first alphabetically -- elogind-dinit over
elogind-openrc -- pulling in dinit -> dinit-rc, which conflicts with
openrc (pulled by elogind-openrc via dbus-openrc).
Fix: use operations.insert(0, ...) so elogind-{init} installs as a
separate pacman call BEFORE base, satisfying init-logind before base
needs it.
- Add antergos-next-keyring to basestrap install list so keyring files
are available in the target early (uses host keyring which already
trusts the Antergos key)
- Add antergos-next to keyrings list so pacman-key --populate imports
the Antergos key into the target's keychain
- try_remove artix-release so antergos-release doesn't conflict when
packages module runs later (replaces= PKGBUILD fix also pushed, but
this works regardless of rebuild)
- Fix remove() method in basestrap main.py: was missing --root <target>,
causing removals to operate on the HOST instead of the target
The basestrap module from calamares-extensions had a hardcoded check that
skipped installing 'base' if hasInternet wasn't True in global storage.
This silently left the target empty, causing the packages module to fail
with error 127 (pacman not found in chroot).
Changes:
- Make hasInternet check configurable via skip_if_no_internet (default false),
matching the packages module pattern
- Fix potential NameError when netinstallAdd isn't in global storage
- Copy basestrap.conf to global modules override so it's available to online
mode (the launcher only copies settings.conf, not module configs)
c-ludenberg