[Bug]: Kernel Manager skips essential pre-build steps — no _localmodcfg_path resolution, no GPG key handling

[AndiB74]

Package name

cachyos-kernel-manager

Package version

1.19.0

Issue type

Build failure / incorrect behavior

Description

Two related gaps, both confirmed by grepping the source (develop branch),
where Kernel Manager skips preparation steps that a build genuinely needs
and instead lets makepkg fail with a raw, unhelpful error 30+ minutes
into a build (or immediately, in the GPG case) — rather than validating
upfront.

---

1. "Modprobed-db" option never sets/reads _localmodcfg_path

The checkbox only toggles _localmodcfg=yes in the PKGBUILD. It never
sets, reads, or validates _localmodcfg_path:

$ grep -rn "localmodcfg" src/
src/config-options.hpp:34:    bool localmodcfg_check{};
src/config-options.cpp:61:        .localmodcfg_check = rust_config_options.localmodcfg_check,
src/config-options.cpp:88:        .localmodcfg_check = config_options.localmodcfg_check,
src/conf-window.cpp:171:    {&Ui::ConfOptionsPage::localmodcfg_check, &ConfigOptions::localmodcfg_check, "localmodcfg"},
src/compile_options.json:6:        "localmodcfg": "_localmodcfg",

No code anywhere references _localmodcfg_path, modprobed-db.conf, or
DBPATH. Kernel Manager relies entirely on the PKGBUILD's own default:

: "${_localmodcfg_path:="$HOME/.config/modprobed.db"}"

modprobed-db itself is configured via ~/.config/modprobed-db.conf
(see its manpage), where DBPATH can point anywhere the user has set it.
When DBPATH doesn't match the PKGBUILD's hardcoded default, Kernel
Manager builds with a stale, empty, or wrong module list — with no
warning. Depending on whether a stale file happens to exist at the old
default path, this results in either:

• ==> ERROR: No modprobed.db data found (build aborts), or
• a build that silently drops boot-critical/hardware-critical modules
  (WiFi, audio, storage drivers) because it read an empty/stale db.

Independently reported by two users, same symptom, different root causes:

• [Bug]: _localmodcfg needs an absolute path CachyOS-PKGBUILDS#1853
• https://discuss.cachyos.org/t/the-kernel-manager-does-not-see-the-modeprobed-db-database/35671

Suggested fix: parse ~/.config/modprobed-db.conf for DBPATH and pass
_localmodcfg_path=$DBPATH/modprobed.db explicitly as an absolute path
(makepkg's build environment does not reliably expand ~/$HOME). At
minimum, warn in the UI if the resolved path doesn't exist before
starting the build.

---

2. No GPG/PGP key handling before invoking makepkg

$ grep -rn -i "gpg\|pgp\|recv-keys\|validpgpkeys\|keyserver" src/
(no results)

$ grep -rn "makepkg" src/
src/aur_kernel.cpp:53:   utils::runCmdTerminal("makepkg -sicf --cleanbuild --skipchecksums", false);
src/conf-window.cpp:734: run_cmd_async("makepkg -scf --cleanbuild --skipchecksums && touch .done-status", ...);

No --skippgpcheck, no pre-check of validpgpkeys= from the PKGBUILD,
no automatic import of missing keys, and no try/catch or error surfacing
around the makepkg call in aur_kernel.cpp. If a required signing key
isn't already in the user's local GPG keyring, the build simply fails on
makepkg's own PGP verification with a raw error the user has to
interpret and fix manually (gpg --recv-keys ...) outside the tool.

Reproduced on a fresh CachyOS install (X99 board, i7-58xx, GTX 1070) —
first build attempt failed on an unimported PGP key with no guidance
from the UI on what to do next.

Suggested fix: before invoking makepkg, extract validpgpkeys= from the
PKGBUILD, check each against the local keyring (gpg --list-keys), and
auto-import any missing ones from a keyserver (with a fallback server)
— same approach already working reliably in third-party build scripts,
e.g.:

for key_id in $key_ids
    if gpg --list-keys $key_id &>/dev/null
        # already present
    else
        gpg --keyserver hkps://keys.openpgp.org --recv-keys $key_id \
            || gpg --keyserver hkps://pgp.surf.nl --recv-keys $key_id
    end
end

---

This is not a one-off: the same class of failure (unimported PGP key,
no guidance from the tool) has been reported and reproduced by multiple
independent users since June 2026, still without a fix in Kernel Manager
itself — the only working workaround remains running gpg --recv-keys
manually, outside the tool, before building:
https://discuss.cachyos.org/t/cachyos-kernel-manager-fails-on-custom-config/31802

(Note: run without sudo — the key must land in the invoking user's own
keyring, not root's, since makepkg refuses to run as root anyway.)
Both point to the same underlying gap: Kernel Manager hands off to
makepkg without validating the preconditions it needs to succeed,
so failures surface late (sometimes 30+ minutes in) and unhelpfully,
instead of being caught and explained upfront in the GUI.

Steps to reproduce

For (1): See steps in linked issue #1853 / forum thread above.

For (2):

1. Fresh system / fresh GPG keyring, no CachyOS kernel keys imported yet.
2. Open Kernel Manager, select a CachyOS kernel variant, build.
3. Build fails on PGP signature verification with no in-app guidance.

Relevant logs or error output

(attach your own PGP error log excerpt here if you still have it)

CachyOS bug report

No response

Additional context

No response