Conversation
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
package structure Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
fix: addition of functions for understanding the booters Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
fix: removed unnecessary constants Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
forcing --removable --no-nvram, matching systemd-boot/refind's own install tools; update booter.toml's stale rationale comment Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
WrittenBootEntry enum (Bls/Uki) so callers branch on the actual boot-resource kind instead of comparing entry_name against a magic UKI slot constant Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
installers (refind-install/bootctl), matching grub's already-real install() Also wires up install()'s C-ABI entry point for uki/refind/systemd-boot, which was a hardcoded stub always returning 0 without ever calling the backend — plugin.install() goes exclusively through this extern "C" fn in both static and dynamic link modes, so uki's existing register_slot NVRAM logic never actually ran either. booter.toml gains install_bin for both plugins; systemd-boot's build.rs now generates both the shared [boot] and its own [systemd_boot] section like refind already did. refactor: propagate the real ErrorKind through boot-plugin request-conversion errors instead of discarding it to a hardcoded InvalidRequest Adds From<ErrorKind> for GrubError/UkiError/RefindError/SystemdBootError (PermissionDenied passes through, everything else still falls back to InvalidRequest) so try_from(...).map_err(XxxError::from) can reuse it instead of a closure that threw the conversion error away. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
export boundary (setup_existing/setup_whole_disk/setup_abi_version/cancel), matching upac-lib's own export/mod.rs Renames lib/setup/src/genesis/ to stages/ (source.rs->prepare.rs, embed.rs->database.rs, entry.rs->boot.rs), extracts SourceArchive into its own archive.rs, and drops the old genesis/types.rs and data.rs in favor of SetupExistingData/SetupWholeDiskData built directly from the new CSetupExistingRequest/CSetupWholeDiskRequest C-ABI types. Real boot stage: resolves the boot plugin, calls install()/write_boot_entry()/set_one_shot(), UKI to/from-slot seed-copy. Manual-mode ESP partition geometry (partition number, starting/ending LBA, unique GUID) is now auto-detected from the existing GPT via a new partition::existing_esp_geometry() instead of being silently zeroed, so TargetSysroot no longer needs Option-wrapped geometry fields at all. Adds error.rs's missing From<SetupError> for ErrorKind and the crate's [lib] crate-type (cdylib+rlib), so the new extern "C" fns are actually exportable for up-sp to link against, either dynamically or statically. Kernel stage stays a no-op stub — tracked separately Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
WrittenBootEntry return type, matching installer/update/rollback's already-fixed call sites Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
`setup_cancel` Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
C-ABI export functions, mirroring up's own libcore.rs/types::errors
pattern
Reorganizes src/ into upac-cli's types/{mod,errors,progress}.rs shape,
adds libcore.rs (Lib::load() resolving
setup_existing/setup_whole_disk/setup_cancel/setup_abi_version as static
fn pointers — always static-link, genesis has nowhere to dlopen a .so
from on a blank disk) plus a shared invoke() helper.
commands/{auto,manual}.rs (whole_disk.rs renamed to auto.rs) now build
SetupExistingRequest/SetupWholeDiskRequest and cross the C-ABI boundary
directly instead of calling
upac_setup::stages::run_existing/run_whole_disk as plain Rust.
types::errors::LibError/error_kind_message replace the old rich
SetupError Display impl, formatting straight off CError's
domain-agnostic ErrorKind + SetupStateId::from_stage_index — same shape
up already uses, just scoped to one domain. i18n keys swapped to the
generic ErrorKind set to match.
Renames upac-setup's own cancel export to setup_cancel — it collided
with upac-lib's identically-named #[no_mangle] cancel, which also ends
up statically linked into up-sp since upac-setup depends on upac-lib for
its own package-install machinery
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
plugin instead of whichever single one happens to exist Adds BootResourceKind (Bls/Uki, same axis WrittenBootEntry already uses) as a new required Booter trait method plus C-ABI query (boot_resource_kind, BOOT_ABI_VERSION 2->3), implemented by all four booter plugins (uki -> Uki, grub/systemd-boot/refind -> Bls) and threaded through BootPlugin/static_link.rs/dynamic_link.rs. write_boot_entry now takes the wanted kind and filters get_boot_resources() against it — errors UnsupportedBootResource if the wanted kind is absent even when a different kind is present, AmbiguousBootResource only among matches of the wanted kind, NoBootResource only when the tree has nothing at all. Reorders all 5 ordinary checkout.rs stages (installer/update/rollback/files/uninstaller) to resolve the boot plugin before calling write_boot_entry, since the call now needs plugin.boot_resource_kind() as an argument — genesis's own stages/boot.rs already had the right order. Closes the two TODO.md items about write_boot_entry's resource-selection blindness and the checkout.rs call ordering Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
(cancel/version_abi never got renamed to lib_cancel/lib_abi_version)
The static-link path referenced upac::export::{cancel, version_abi} and
upac_abi::ABI_VERSION, none of which exist anymore — upac-lib's own
export/mod.rs already uses lib_cancel/lib_abi_version/LIB_ABI_VERSION,
but the dynamic-plugins dlopen path (which resolves symbols by string
name) had the same stale "cancel"/"version_abi" names too. Both paths
now agree with what upac-lib actually exports; `cargo clippy -p upac-cli
-p upac-lib --no-default-features --features
upac-cli/builtin-all,upac-lib/builtin-all` (the CI "fully static" build)
is clean again
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
earlier refactors
database.rs/orchestrator.rs/scripts_hook.rs imported types from their
pre-refactor locations
(upac_types::{DeclarativeTrigger,FileEntry,FileEntryScope,PackageMeta}
at crate root instead of their actual submodules; ProgressEventBuilder
from upac_abi::hook instead of upac_types::hook;
scripts::load::load_hooks instead of scripts::load_hooks).
plugin_boot_error.rs asserted a BootPluginError::AmbiguousClaim variant
that no longer exists. plugin_boot_manifest.rs and half of
plugin_decoder.rs tested
load_boot_plugin_manifests/load_decoder_manifests as
directory+extension-parameterized free functions — both are now
BootPluginManifests::new()/DecoderManifests::new(), which read
unconditionally from a hardcoded lib.toml-configured path with no way to
inject a test directory, so that style of test isn't reconstructible;
plugin_boot_manifest.rs is deleted and plugin_decoder.rs keeps only its
still-valid build_trigger_table coverage. `cargo test -p upac-lib
--all-targets` is green
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
converted tests to use inline definitions within the module, following convention fix: removed the unnecessary CLI structure and replaced it with an enum fix: renamed folders to use single-word names new: now can build composefs-setup-root bin as command Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RustToC/CTryToRust assume any field type not listed in SHARED_TYPES has a separate CXxx composite counterpart to convert through. InitramfsGenerator is a plain Copy enum used directly on both sides, like FsKind already is — add it to SHARED_TYPES so the derives stop looking for a nonexistent CInitramfsGenerator Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Setup requests hardcoded dracut as the only initramfs generator with no way to ask for mkinitcpio instead. Add a plain repr(u8) InitramfsGenerator enum (Dracut/Mkinitcpio) and thread it through CSetupExistingRequest/ CSetupWholeDiskRequest Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Both auto and manual mode were passing a hardcoded Dracut regardless of user intent. Add a clap ValueEnum wrapper (InitramfsGeneratorClapArg, needed for the same orphan-rule reason FsKind already has one) and a --initramfs-generator flag whose default comes from cli.toml's new [initramfs] section, mirroring how --deploy-fs already defaults from disk_defaults Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Add initramfs_generator to SetupExistingRequest/SetupWholeDiskRequest so the new ABI field actually reaches the domain layer. Also fix SetupStateId's variant order, which had drifted from the real stage pipeline (Kernel now runs before EmbedDatabase, not after) and was silently misreporting which stage failed Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
KernelStage needs a real directory on disk to hand to dracut/mkinitcpio, but only the tree-to-disk direction (import_directory) existed. Add export_directory plus small per-node-kind helpers (export_leaf/ export_symlink/export_regular_file, and a shared regular_file_content also reused by read_file) so the recursive walk stays a thin dispatcher instead of inlining the content-resolution match Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Detect the kernel version from the imported tree, export the whole prefix tree to a scratch directory, and run dracut or mkinitcpio (non-hostonly/generic flags, since genesis targets an arbitrary disk, not the build host) to produce initramfs.img/uki.efi, then import the result back under lib/modules/<kver>/. Output filenames move to lib.toml's [genesis] section rather than being hardcoded. Requires KernelStage to run before EmbedDatabaseStage, since committing the tree removes PrefixTree from context Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
genesis's system/ import mechanism expects lib/systemd/system/ composefs-setup-root.service in the source tree, but that alone isn't enough — the initrd needs its own module-setup.sh to pull the binary and unit into the generated initramfs and wire the initrd-root-fs.target enablement (systemd running inside the initrd never sees anything add-wants'd into the real root's own unit tree after switch-root). Sourced from containers/composefs-rs's own dracut module example (MIT OR Apache-2.0) rather than hand-rolled, since dracut module ordering/conditions are easy to get subtly wrong Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
initrd-root-fs.target only exists inside the initrd's own systemd
instance, so a *.target.wants/ symlink written into the real root tree
is a no-op after switch-root — the real root's systemd never processes
it. The dracut module (hooks/dracut/37composefs/) already creates the
actual enablement at initrd-build time via `$SYSTEMCTL --root
"${initdir}" add-wants`, which is the only place it can take effect
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Unlike dracut, mkinitcpio doesn't force systemd inside the initramfs — confirmed by reading its actual /init (plain busybox ash: run_hook → mount_handler → run_latehook → switch_root) and composefs-setup-root's own source (pure rustix syscalls, no systemd dependency at all beyond a best-effort /run/systemd/volatile-root workaround that no-ops cleanly without it). Add install/composefs + hooks/composefs mirroring mkinitcpio's own two-file hook convention (install/ copies the binary and runscript at build time, hooks/ runs run_latehook() after the real root device is mounted, before switch_root) — no flags needed, since composefs-setup-root's --sysroot default already matches mkinitcpio's own /sysroot convention. Also fixes REUSE compliance for the two vendored dracut files: they carry their own upstream license text as prose, not machine-readable SPDX tags, so reuse lint flagged them as unlicensed. Annotate them externally in REUSE.toml instead of injecting tags into third-party code, and vendor the newly-referenced license texts (MIT, Apache-2.0, LGPL-2.1-or-later) via `reuse download --all` Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Key changes: