path: root/tests/unit/test_archangel.bats

#!/usr/bin/env bats
# Unit tests for the installer/archangel monolith.
#
# Coverage scope: gather_input() in unattended mode — defaulting of
# optional values, preservation of explicit ones, and the
# filesystem-specific encryption checks. Required-field, disk, and
# timezone validation moved to validate_config (called from main
# before gather_input); its coverage lives in test_config.bats.
# The interactive branch (everything reachable via
# `if [[ "$UNATTENDED" != true ]]`) is not unit-tested per the
# project's testing-strategy.org policy on fzf / arch-chroot /
# mkfs / cryptsetup wrappers.
#
# Sourcing archangel relies on the source-guard at the bottom of
# the script: when sourced, function definitions load but main is
# not called, init_logging is not run (so /tmp/archangel-*.log is
# not created), and the banner is not printed.

setup() {
    # shellcheck disable=SC1091
    source "${BATS_TEST_DIRNAME}/../../installer/archangel"
    UNATTENDED=true
    # Tests that call install_failure_cleanup in this process would
    # otherwise run the real disarm_failure_trap, which clears bats' own
    # EXIT trap, and bats reports a failing assertion from that trap: the
    # failure would vanish instead of printing "not ok". The trap arming
    # itself is tested in a child bash, which sources the real functions.
    disarm_failure_trap() { :; }
    # Point the LUKS-mapping check at an empty directory, so a machine with
    # a real /dev/mapper/cryptroot can't leak into the cleanup tests.
    MAPPER_DIR="$BATS_TEST_TMPDIR/mapper"
    mkdir -p "$MAPPER_DIR"
}

#############################
# Optional-field defaults
#############################
# Default values themselves are pinned in test_config.bats (config.sh
# is the single source of truth). The remaining test here covers the
# adjacent guarantee: gather_input doesn't clobber values the user set.

@test "gather_input unattended preserves explicit non-default values" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda)
    FILESYSTEM=btrfs
    NO_ENCRYPT=yes
    LOCALE="en_GB.UTF-8"
    KEYMAP="dvorak"
    ENABLE_SSH="no"
    gather_input >/dev/null
    [ "$FILESYSTEM" = "btrfs" ]
    [ "$LOCALE" = "en_GB.UTF-8" ]
    [ "$KEYMAP" = "dvorak" ]
    [ "$ENABLE_SSH" = "no" ]
}

#############################
# Filesystem-specific encryption validation
#############################

@test "gather_input unattended errors when ZFS without ZFS_PASSPHRASE and encryption on" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda)
    FILESYSTEM=zfs
    NO_ENCRYPT=no
    ZFS_PASSPHRASE=""
    run gather_input
    [ "$status" -eq 1 ]
    [[ "$output" == *"ZFS_PASSPHRASE"* ]]
}

@test "gather_input unattended errors when Btrfs without LUKS_PASSPHRASE and encryption on" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda)
    FILESYSTEM=btrfs
    NO_ENCRYPT=no
    LUKS_PASSPHRASE=""
    run gather_input
    [ "$status" -eq 1 ]
    [[ "$output" == *"LUKS_PASSPHRASE"* ]]
}

@test "gather_input unattended accepts ZFS with NO_ENCRYPT=yes and no passphrase" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda)
    FILESYSTEM=zfs
    NO_ENCRYPT=yes
    ZFS_PASSPHRASE=""
    run gather_input
    [ "$status" -eq 0 ]
}

#############################
# Filesystem validity
#############################
# Validation moved to validate_filesystem in lib/config.sh — covered
# by test_config.bats. main() calls it between check_config and
# gather_input so a bad FILESYSTEM= never reaches install time.

#############################
# RAID-level defaulting
#############################

@test "gather_input unattended defaults RAID_LEVEL to mirror for multi-disk install" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda /dev/sdb)
    FILESYSTEM=zfs
    NO_ENCRYPT=yes
    RAID_LEVEL=""
    gather_input >/dev/null
    [ "$RAID_LEVEL" = "mirror" ]
}

@test "gather_input unattended preserves an explicit RAID_LEVEL on multi-disk install" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda /dev/sdb /dev/sdc)
    FILESYSTEM=zfs
    NO_ENCRYPT=yes
    RAID_LEVEL="raidz1"
    gather_input >/dev/null
    [ "$RAID_LEVEL" = "raidz1" ]
}

@test "gather_input unattended leaves RAID_LEVEL empty for single-disk install" {
    HOSTNAME=h
    TIMEZONE=UTC
    ROOT_PASSWORD=x
    SELECTED_DISKS=(/dev/sda)
    FILESYSTEM=zfs
    NO_ENCRYPT=yes
    RAID_LEVEL=""
    gather_input >/dev/null
    [ -z "$RAID_LEVEL" ]
}

#############################
# Failure trap arming
#############################
# arm_failure_trap / disarm_failure_trap own the trap set that routes a
# mid-install failure to install_failure_cleanup. These run in a child
# bash (via `run bash -c`) because trap inheritance is a property of
# the live shell. Before the EXIT trap, a failure inside any install step
# ended the script through errexit without the cleanup running, and an
# `exit 1` from error() reached no ERR trap at all. Both were how the
# 2026-08-06 and 2026-09-12 retries found a still-mounted disk.

@test "arm_failure_trap fires the cleanup for a command failing inside a function" {
    run bash -c 'source "$1"
        install_failure_cleanup() { disarm_failure_trap; echo CLEANUP-RAN; exit 7; }
        arm_failure_trap
        f() { false; }
        f
        echo NOT-REACHED' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 7 ]
    [[ "$output" == *"CLEANUP-RAN"* ]]
    [[ "$output" != *"NOT-REACHED"* ]]
}

@test "arm_failure_trap fires the cleanup when error() exits from inside a function" {
    run bash -c 'source "$1"
        install_failure_cleanup() { disarm_failure_trap; echo CLEANUP-RAN; exit 7; }
        arm_failure_trap
        f() { false || error "pacstrap failed"; }
        f
        echo NOT-REACHED' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 7 ]
    [[ "$output" == *"CLEANUP-RAN"* ]]
    [[ "$output" != *"NOT-REACHED"* ]]
}

@test "arm_failure_trap fires the cleanup for a failing pipeline inside a function" {
    run bash -c 'source "$1"
        install_failure_cleanup() { disarm_failure_trap; echo CLEANUP-RAN; exit 7; }
        arm_failure_trap
        f() { printf "\n" | false; }
        f
        echo NOT-REACHED' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 7 ]
    [[ "$output" == *"CLEANUP-RAN"* ]]
    [[ "$output" != *"NOT-REACHED"* ]]
}

@test "disarm_failure_trap lets the success path exit without running the cleanup" {
    run bash -c 'source "$1"
        install_failure_cleanup() { disarm_failure_trap; echo CLEANUP-RAN; exit 7; }
        arm_failure_trap
        f() { true; }
        f
        disarm_failure_trap
        echo SUCCESS' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 0 ]
    [[ "$output" == *"SUCCESS"* ]]
    [[ "$output" != *"CLEANUP-RAN"* ]]
}

@test "arm_failure_trap fires the cleanup when the install is sent TERM" {
    run bash -c 'source "$1"
        install_failure_cleanup() { disarm_failure_trap; echo CLEANUP-RAN; exit 7; }
        arm_failure_trap
        f() { kill -TERM $$; sleep 2; echo NOT-REACHED; }
        f' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 7 ]
    [[ "$output" == *"CLEANUP-RAN"* ]]
    [[ "$output" != *"NOT-REACHED"* ]]
}

# A failing command substitution whose status is masked (an argument to
# echo, a `local` declaration, a `|| true`) is not an install failure. With
# errtrace on, bash ran the whole cleanup inside the substitution's
# subshell, unmounting /mnt while the install carried on in the parent.
@test "a masked failing command substitution does not run the cleanup" {
    local marker="$BATS_TEST_TMPDIR/cleanup-ran"
    run bash -c 'source "$1"; M="$2"
        install_failure_cleanup() { disarm_failure_trap; echo ran >> "$M"; exit 7; }
        arm_failure_trap
        f() { echo "$(false)" >/dev/null; local y; y=$(false) || true; echo STEP-DONE; }
        f
        disarm_failure_trap
        echo SUCCESS' _ "${BATS_TEST_DIRNAME}/../../installer/archangel" "$marker"
    [ "$status" -eq 0 ]
    [[ "$output" == *"STEP-DONE"* ]]
    [[ "$output" == *"SUCCESS"* ]]
    [ ! -e "$marker" ]
}

# A SIGTERM to the whole process group also reaches the tee init_logging
# puts on stdout. With the tee gone, the cleanup's first warn raised
# SIGPIPE and killed the shell before anything was unmounted: a Btrfs
# install TERM'd mid-pacstrap in the VM on 2026-09-13 was left with /mnt
# and the LUKS mapping still open. Ctrl-C's SIGINT leaves the tee alive
# (bash starts a process substitution with SIGINT ignored), so the INT
# test guards the ordinary Ctrl-C path rather than the SIGPIPE fix.
#
# The child starts through `env --default-signal=INT` because bats runs it
# in the background, and a background job in a non-interactive shell
# begins with SIGINT ignored, which bash then refuses to trap. Without the
# reset, the INT case would pass by never delivering the signal at all.
signal_group_with_logging_tee() {
    local sig="$1" d="$2"
    env --default-signal=INT setsid bash -c 'source "$1"; D="$2"
        FILESYSTEM=zfs; POOL_NAME=zroot
        umount() { echo "umount $*" >> "$D/calls"; }
        zpool() { echo "zpool $*" >> "$D/calls"; [[ "$1" == list ]]; }
        exec > >(tee -a "$D/tee.log") 2>&1
        arm_failure_trap
        grep SigIgn /proc/$$/status > "$D/sigign"
        echo $$ > "$D/armed"
        sleep 30' _ "${BATS_TEST_DIRNAME}/../../installer/archangel" "$d" </dev/null >/dev/null 2>&1 &

    local i pid=""
    for i in $(seq 1 100); do
        [ -s "$d/armed" ] && pid=$(cat "$d/armed") && break
        sleep 0.1
    done
    [ -n "$pid" ] || return 1
    kill "-$sig" -- "-$pid"
    for i in $(seq 1 100); do
        kill -0 "$pid" 2>/dev/null || break
        sleep 0.1
    done
    kill -0 "$pid" 2>/dev/null && kill -KILL -- "-$pid"
    return 0
}

@test "install_failure_cleanup finishes when SIGTERM to the group also killed its logging tee" {
    local d="$BATS_TEST_TMPDIR"
    signal_group_with_logging_tee TERM "$d"
    grep -qx "zpool export zroot" "$d/calls"
}

@test "install_failure_cleanup finishes when Ctrl-C's SIGINT reaches the whole group" {
    local d="$BATS_TEST_TMPDIR"
    signal_group_with_logging_tee INT "$d"
    # Guard against the vacuous pass: SIGINT (mask bit 0x2) must not be
    # ignored in the child, or the signal never arrived.
    local mask
    mask=$(awk '{print $2}' "$d/sigign")
    [ $(( 0x$mask & 0x2 )) -eq 0 ]
    grep -qx "zpool export zroot" "$d/calls"
}

# The busy retries can keep the cleanup silent for several seconds, and a
# second Ctrl-C or SIGTERM in that window used to kill it halfway: the
# export never finished and the pool stayed imported.
@test "a second SIGTERM during the busy-retry window doesn't stop the cleanup" {
    local d="$BATS_TEST_TMPDIR"
    env --default-signal=INT setsid bash -c 'source "$1"; D="$2"
        FILESYSTEM=zfs; POOL_NAME=zroot; CLEANUP_BUSY_ATTEMPTS=6
        umount() { :; }
        zpool() {
            [[ "$1" == list ]] && return 0
            if [[ "$*" == "export zroot" ]]; then
                echo try >> "$D/exports"
                [ "$(wc -l < "$D/exports")" -ge 4 ] && { echo exported >> "$D/done"; return 0; }
                return 1
            fi
            [[ "$*" == "export -f zroot" ]] && echo forced >> "$D/done"
            return 0
        }
        arm_failure_trap
        echo $$ > "$D/armed"
        sleep 30' _ "${BATS_TEST_DIRNAME}/../../installer/archangel" "$d" </dev/null >/dev/null 2>&1 &

    local i pid=""
    for i in $(seq 1 100); do
        [ -s "$d/armed" ] && pid=$(cat "$d/armed") && break
        sleep 0.1
    done
    [ -n "$pid" ]
    kill -TERM -- "-$pid"
    for i in $(seq 1 50); do
        [ -s "$d/exports" ] && break
        sleep 0.1
    done
    [ -s "$d/exports" ]
    kill -TERM -- "-$pid" 2>/dev/null || true
    for i in $(seq 1 100); do
        kill -0 "$pid" 2>/dev/null || break
        sleep 0.1
    done
    kill -0 "$pid" 2>/dev/null && kill -KILL -- "-$pid"

    [ "$(cat "$d/done" 2>/dev/null)" = "exported" ]
}

@test "install_failure_cleanup runs exactly once when its own exit re-fires the trap" {
    run bash -c 'source "$1"
        FILESYSTEM=zfs; POOL_NAME=zroot
        umount() { :; }
        zpool() { return 1; }
        arm_failure_trap
        f() { false; }
        f' _ "${BATS_TEST_DIRNAME}/../../installer/archangel"
    [ "$status" -eq 1 ]
    [ "$(grep -c 'cleaning up' <<<"$output")" -eq 1 ]
    [[ "$output" == *"system cleaned up"* ]]
}

#############################
# install_failure_cleanup
#############################
# install_failure_cleanup is the trap target for ERR / INT / TERM / EXIT
# during install_zfs and install_btrfs. It clears sensitive vars,
# dispatches on FILESYSTEM, and exits non-zero. Tests use function
# overrides to capture which system tools the cleanup invokes; the
# tools themselves (umount, zpool, btrfs_cleanup,
# btrfs_close_encryption) are deliberately VM-tested per
# testing-strategy.org.

#############################
# retry_busy
#############################
# Teardown after an interrupted pacstrap races pacman's children in the
# chroot, which can still be exiting: the first export or LUKS close fails
# busy and succeeds a moment later. retry_busy gives it a bounded window.

@test "retry_busy runs a command that succeeds once, without sleeping" {
    CALLS=()
    sleep() { CALLS+=("sleep $*"); }
    ok() { CALLS+=("ok"); return 0; }
    retry_busy 5 ok
    [ "${#CALLS[@]}" -eq 1 ]
    [ "${CALLS[0]}" = "ok" ]
}

@test "retry_busy succeeds when the command first succeeds on the last attempt" {
    CALLS=(); TRIES=0
    sleep() { CALLS+=("sleep $*"); }
    flaky() { TRIES=$((TRIES + 1)); [ "$TRIES" -ge 3 ]; }
    retry_busy 3 flaky
    [ "$TRIES" -eq 3 ]
    [ "${#CALLS[@]}" -eq 2 ]
}

@test "retry_busy returns 1 after the last attempt, sleeping only between attempts" {
    CALLS=(); TRIES=0
    sleep() { CALLS+=("sleep $*"); }
    never() { TRIES=$((TRIES + 1)); return 1; }
    run retry_busy 4 never
    [ "$status" -eq 1 ]
    retry_busy 4 never || true
    [ "$TRIES" -eq 4 ]
    [ "${#CALLS[@]}" -eq 3 ]
}

# bats fails a test through errexit. The cleanup turns errexit off for
# itself, and before `local -` scoped that, every assertion after a direct
# call except the last one stopped being able to fail.
@test "install_failure_cleanup leaves the caller's errexit on" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    umount() { :; }
    zpool() { return 1; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    # Capture the flags, then restore errexit before asserting. If the
    # cleanup leaked errexit-off, a bare assertion here couldn't fail.
    local flags="$-"
    set -e
    [[ "$flags" == *e* ]]
}

@test "install_failure_cleanup ZFS path retries the export while the pool is still busy" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    CALLS=(); EXPORTS=0
    umount() { :; }
    zpool() {
        CALLS+=("zpool $*")
        [[ "$1" == list ]] && return 0
        if [[ "$*" == "export zroot" ]]; then
            EXPORTS=$((EXPORTS + 1))
            [ "$EXPORTS" -ge 3 ]
            return
        fi
        return 0
    }
    sleep() { CALLS+=("sleep $*"); }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [ "$EXPORTS" -eq 3 ]
    [[ " ${CALLS[*]} " != *" zpool export -f zroot "* ]]
}

@test "install_failure_cleanup ZFS path forces the export once the busy window runs out" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    CLEANUP_BUSY_ATTEMPTS=4
    CALLS=(); EXPORTS=0; SLEEPS=0
    umount() { :; }
    zpool() {
        CALLS+=("zpool $*")
        [[ "$1" == list ]] && return 0
        [[ "$*" == "export zroot" ]] && { EXPORTS=$((EXPORTS + 1)); return 1; }
        return 0
    }
    sleep() { SLEEPS=$((SLEEPS + 1)); }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [ "$EXPORTS" -eq 4 ]
    [ "$SLEEPS" -eq 3 ]
    [[ " ${CALLS[*]} " == *" zpool export -f zroot "* ]]
}

@test "install_failure_cleanup Btrfs path retries closing LUKS until the mapping is gone" {
    FILESYSTEM=btrfs
    CLOSES=0
    umount() { :; }
    btrfs_cleanup() { :; }
    btrfs_close_encryption() { CLOSES=$((CLOSES + 1)); }
    luks_mappings_open() { [ "$CLOSES" -lt 3 ]; }
    sleep() { :; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [ "$CLOSES" -eq 3 ]
}

@test "install_failure_cleanup Btrfs path stops retrying the LUKS close after the busy window" {
    FILESYSTEM=btrfs
    CLEANUP_BUSY_ATTEMPTS=4
    CLOSES=0
    umount() { :; }
    btrfs_cleanup() { :; }
    btrfs_close_encryption() { CLOSES=$((CLOSES + 1)); }
    luks_mappings_open() { return 0; }
    sleep() { :; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [ "$CLOSES" -eq 4 ]
}

@test "install_failure_cleanup clears sensitive variables before exiting" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    ROOT_PASSWORD="topsecret"
    ZFS_PASSPHRASE="anothersecret"
    LUKS_PASSPHRASE="thirdsecret"

    # Mocks: silent no-ops for system tools; error returns non-zero
    # so the function returns instead of exiting the test process.
    umount() { :; }
    zpool() { return 1; }
    btrfs_cleanup() { :; }
    btrfs_close_encryption() { :; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [ -z "$ROOT_PASSWORD" ]
    [ -z "$ZFS_PASSPHRASE" ]
    [ -z "$LUKS_PASSPHRASE" ]
}

@test "install_failure_cleanup dispatches to ZFS path when FILESYSTEM=zfs" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    CALLS=()

    # Mocks track invocations via CALLS array. Array assignment is not
    # affected by the production code's >/dev/null 2>&1 redirects on
    # the zpool list check, so we capture the call regardless of where
    # the mock's stdout would have gone.
    umount() { CALLS+=("umount $*"); return 0; }
    zpool() {
        CALLS+=("zpool $*")
        [[ "$1" == "list" ]] && return 0
        return 0
    }
    btrfs_cleanup() { CALLS+=("btrfs_cleanup"); }
    btrfs_close_encryption() { CALLS+=("btrfs_close_encryption"); }
    warn() { :; }
    error() { CALLS+=("error"); return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" umount /mnt/efi "* ]]
    [[ " ${CALLS[*]} " == *" umount -R /mnt "* ]]
    [[ " ${CALLS[*]} " == *" zpool list zroot "* ]]
    [[ " ${CALLS[*]} " == *" zpool export zroot "* ]]
    [[ " ${CALLS[*]} " != *" btrfs_cleanup "* ]]
    [[ " ${CALLS[*]} " != *" btrfs_close_encryption "* ]]
}

@test "install_failure_cleanup dispatches to Btrfs path when FILESYSTEM=btrfs" {
    FILESYSTEM=btrfs
    CALLS=()

    umount() { CALLS+=("umount $*"); return 0; }
    zpool() { CALLS+=("zpool $*"); return 0; }
    btrfs_cleanup() { CALLS+=("btrfs_cleanup"); }
    btrfs_close_encryption() { CALLS+=("btrfs_close_encryption"); }
    warn() { :; }
    error() { CALLS+=("error"); return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" umount /mnt/efi "* ]]
    [[ " ${CALLS[*]} " == *" btrfs_cleanup "* ]]
    [[ " ${CALLS[*]} " == *" btrfs_close_encryption "* ]]
    [[ " ${CALLS[*]} " != *" zpool"* ]]
}

# btrfs_cleanup unmounts only the subvolumes it mounted, one level at a
# time. A pacstrap interrupted mid-transaction can leave its /proc, /sys
# and /dev bind mounts under /mnt, which keep the root busy, so the LUKS
# mapping can't close and the retry's disk_in_use still sees a mountpoint.
@test "install_failure_cleanup Btrfs path unmounts the target root recursively before closing LUKS" {
    FILESYSTEM=btrfs
    CALLS=()

    umount() { CALLS+=("umount $*"); return 0; }
    zpool() { CALLS+=("zpool $*"); return 0; }
    btrfs_cleanup() { CALLS+=("btrfs_cleanup"); }
    btrfs_close_encryption() { CALLS+=("btrfs_close_encryption"); }
    warn() { :; }
    error() { CALLS+=("error"); return 1; }

    install_failure_cleanup || true

    local i recursive=-1 close=-1
    for i in "${!CALLS[@]}"; do
        [[ "${CALLS[$i]}" == "umount -R /mnt" ]] && recursive=$i
        [[ "${CALLS[$i]}" == "btrfs_close_encryption" ]] && close=$i
    done
    [ "$recursive" -ge 0 ]
    [ "$close" -gt "$recursive" ]
}

@test "install_failure_cleanup Btrfs path falls back to a lazy recursive unmount when the root is busy" {
    FILESYSTEM=btrfs
    CALLS=()

    umount() {
        CALLS+=("umount $*")
        [[ "$*" == "-R /mnt" ]] && return 32
        return 0
    }
    btrfs_cleanup() { CALLS+=("btrfs_cleanup"); }
    btrfs_close_encryption() { CALLS+=("btrfs_close_encryption"); }
    warn() { :; }
    error() { CALLS+=("error"); return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" umount -R -l /mnt "* ]]
    [[ " ${CALLS[*]} " == *" btrfs_close_encryption "* ]]
}

@test "install_failure_cleanup ZFS path skips zpool export when pool not imported" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    CALLS=()

    umount() { CALLS+=("umount $*"); return 0; }
    zpool() {
        CALLS+=("zpool $*")
        [[ "$1" == "list" ]] && return 1  # pool NOT imported
        return 0
    }
    btrfs_cleanup() { :; }
    btrfs_close_encryption() { :; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" zpool list zroot "* ]]
    [[ " ${CALLS[*]} " != *" zpool export"* ]]
}

@test "install_failure_cleanup ZFS path falls back to lazy unmount when a mount is busy" {
    FILESYSTEM=zfs
    POOL_NAME=zroot
    CALLS=()

    # A pacstrap-interrupted target can leave busy mounts that a plain
    # umount can't release; cleanup must retry lazily so the retry sees a
    # clean disk. Non-lazy umount fails here; the -l fallback succeeds.
    umount() {
        CALLS+=("umount $*")
        [[ "$*" == *"-l"* ]] && return 0
        return 1
    }
    zpool() { CALLS+=("zpool $*"); return 0; }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" umount -l /mnt/efi "* ]]
    [[ " ${CALLS[*]} " == *" umount -R -l /mnt "* ]]
    # The pool still gets exported after the lazy unmount.
    [[ " ${CALLS[*]} " == *" zpool export zroot "* ]]
}

@test "install_failure_cleanup Btrfs path falls back to lazy unmount when EFI is busy" {
    FILESYSTEM=btrfs
    CALLS=()

    umount() {
        CALLS+=("umount $*")
        [[ "$*" == *"-l"* ]] && return 0
        return 1
    }
    btrfs_cleanup() { CALLS+=("btrfs_cleanup"); }
    btrfs_close_encryption() { CALLS+=("btrfs_close_encryption"); }
    warn() { :; }
    error() { return 1; }

    install_failure_cleanup || true

    [[ " ${CALLS[*]} " == *" umount -l /mnt/efi "* ]]
}

#############################
# validate_environment
#############################
# Boundary wrappers (is_uefi_boot, required_commands) are stubbed so the
# composition's fail-fast wiring is exercised without depending on the
# host's firmware mode or installed tools. The real command list lives in
# test_common.bats; the real UEFI/network probes run in the VM harness.

@test "validate_environment errors when not booted in UEFI mode" {
    is_uefi_boot() { return 1; }
    required_commands() { return 0; }
    FILESYSTEM=zfs
    run validate_environment
    [ "$status" -eq 1 ]
    [[ "$output" == *"UEFI"* ]]
}

@test "validate_environment errors when a required command is missing" {
    is_uefi_boot() { return 0; }
    required_commands() { echo "definitely-not-a-real-cmd-xyz"; }
    FILESYSTEM=zfs
    run validate_environment
    [ "$status" -eq 1 ]
    [[ "$output" == *"definitely-not-a-real-cmd-xyz"* ]]
}

@test "validate_environment passes when UEFI present and commands resolve" {
    is_uefi_boot() { return 0; }
    required_commands() { echo "bash"; }
    FILESYSTEM=zfs
    run validate_environment
    [ "$status" -eq 0 ]
}

#############################
# validate_install_targets
#############################
# disk_in_use / disk_size_bytes / network_available are the system-boundary
# wrappers; stubbing them drives the real composition + real
# disk_meets_min_size. Live probes run in the VM harness on the happy path.

@test "validate_install_targets errors when a disk is in use" {
    SELECTED_DISKS=(/dev/sda)
    disk_in_use() { return 0; }
    disk_size_bytes() { echo 500107862016; }
    network_available() { return 0; }
    run validate_install_targets
    [ "$status" -eq 1 ]
    [[ "$output" == *"in use"* ]]
}

@test "validate_install_targets errors when a disk is too small" {
    SELECTED_DISKS=(/dev/sda)
    disk_in_use() { return 1; }
    disk_size_bytes() { echo 1000000; }
    network_available() { return 0; }
    run validate_install_targets
    [ "$status" -eq 1 ]
    [[ "$output" == *"too small"* ]]
}

@test "validate_install_targets errors when disk size is unreadable" {
    SELECTED_DISKS=(/dev/sda)
    disk_in_use() { return 1; }
    disk_size_bytes() { echo ""; }
    network_available() { return 0; }
    run validate_install_targets
    [ "$status" -eq 1 ]
}

@test "validate_install_targets errors when the network is unreachable" {
    SELECTED_DISKS=(/dev/sda)
    disk_in_use() { return 1; }
    disk_size_bytes() { echo 500107862016; }
    network_available() { return 1; }
    run validate_install_targets
    [ "$status" -eq 1 ]
    [[ "$output" == *"network"* || "$output" == *"connectivity"* ]]
}

@test "validate_install_targets passes when disks idle, large enough, network up" {
    SELECTED_DISKS=(/dev/sda /dev/sdb)
    disk_in_use() { return 1; }
    disk_size_bytes() { echo 500107862016; }
    network_available() { return 0; }
    run validate_install_targets
    [ "$status" -eq 0 ]
}

#############################
# network_available
#############################
# The two probes it composes — DNS via getent and a TCP-443 open via
# timeout+bash /dev/tcp — are the system boundary; mocking them drives the
# fail-fast wiring without a live network. The real probe runs in the VM
# harness. These pin the "network failure before pacstrap" error path that
# validate_install_targets surfaces.

@test "network_available returns 1 when DNS resolution fails" {
    getent() { return 1; }
    run network_available
    [ "$status" -eq 1 ]
}

@test "network_available returns 1 when DNS resolves but the TCP connect fails" {
    getent() { return 0; }
    timeout() { return 1; }
    run network_available
    [ "$status" -eq 1 ]
}

@test "network_available returns 0 when DNS resolves and the TCP connect opens" {
    getent() { return 0; }
    timeout() { return 0; }
    run network_available
    [ "$status" -eq 0 ]
}

#############################
# configure_zfs_keyfile
#############################
# Encrypted ZFS installs prompt for the same passphrase twice:
# ZFSBootMenu unlocks the pool to read the kernel and initramfs, then
# kexecs, and the key doesn't survive kexec — so the booted initramfs
# re-imports the pool, finds keylocation=prompt, and asks again.
#
# configure_zfs_keyfile closes the second prompt the same way the Btrfs
# path already closes its LUKS equivalent: write the passphrase to a
# keyfile inside the encrypted root, point the encryption root at it,
# and bake it into the initramfs via FILES=. ZFSBootMenu can't read a
# file inside a dataset it hasn't unlocked yet, so it still prompts
# once — that surviving prompt is the intended behavior, not a bug.
#
# zfs is the stubbed system boundary. The keyfile write, its
# permissions, and the FILES= wiring are exercised for real.

zfs_keyfile_fixture() {
    TEST_ROOT=$(mktemp -d)
    MNTPOINT="$TEST_ROOT"
    ZFS_ARGS_LOG="$TEST_ROOT/zfs-args"
    mkdir -p "$MNTPOINT/etc"
    printf '%s\n' 'FILES=()' > "$MNTPOINT/etc/mkinitcpio.conf"
    zfs() { echo "$*" >> "$ZFS_ARGS_LOG"; return 0; }
}

@test "configure_zfs_keyfile writes the passphrase with no trailing newline" {
    zfs_keyfile_fixture
    configure_zfs_keyfile "correct horse" zroot
    # The keyfile lands mode 000, which locks out the owner too — only root
    # bypasses that, and these tests don't run as root. Restore read access to
    # inspect the content; the mode itself is asserted separately below.
    chmod u+r "$MNTPOINT/etc/zfs/zroot.key"
    # A trailing newline would become part of the passphrase ZFS reads back,
    # so the key would never match what's typed at the ZBM prompt. 13 bytes,
    # not 14: no terminator.
    [ "$(wc -c < "$MNTPOINT/etc/zfs/zroot.key")" -eq 13 ]
    [ "$(cat "$MNTPOINT/etc/zfs/zroot.key")" = "correct horse" ]
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile points the encryption root at the keyfile" {
    zfs_keyfile_fixture
    configure_zfs_keyfile testpass zroot
    grep -qE '^set +keylocation=file:///etc/zfs/zroot\.key +zroot$' "$ZFS_ARGS_LOG"
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile changes the location without rekeying the pool" {
    zfs_keyfile_fixture
    configure_zfs_keyfile testpass zroot
    # keylocation is settable with plain `zfs set` (zfsprops(7)), and
    # keyformat is already passphrase from pool creation. Reaching for
    # `zfs change-key` here would rekey the pool and prompt for new key
    # material mid-install — and losing keyformat=passphrase would leave
    # ZFSBootMenu with no way to accept a typed passphrase at all.
    ! grep -qF 'change-key' "$ZFS_ARGS_LOG"
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile bakes the keyfile into the initramfs" {
    zfs_keyfile_fixture
    configure_zfs_keyfile testpass zroot
    grep -qF 'FILES=(/etc/zfs/zroot.key)' "$MNTPOINT/etc/mkinitcpio.conf"
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile leaves the keyfile unreadable to other users" {
    zfs_keyfile_fixture
    configure_zfs_keyfile testpass zroot
    # Protected at rest by the encrypted dataset, but a stray mode 644
    # would expose it to any local user on the running system.
    [ "$(stat -c '%a' "$MNTPOINT/etc/zfs/zroot.key")" -eq 0 ]
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile preserves a passphrase containing shell metacharacters" {
    zfs_keyfile_fixture
    configure_zfs_keyfile 'a$b "c" \d*' zroot
    chmod u+r "$MNTPOINT/etc/zfs/zroot.key"
    [ "$(cat "$MNTPOINT/etc/zfs/zroot.key")" = 'a$b "c" \d*' ]
    rm -rf "$TEST_ROOT"
}

@test "configure_zfs_keyfile aborts when the key change fails" {
    zfs_keyfile_fixture
    zfs() { return 1; }
    run configure_zfs_keyfile testpass zroot
    # Silently continuing would ship an initramfs whose keyfile doesn't
    # match the pool, turning one prompt into an unbootable system.
    [ "$status" -eq 1 ]
    rm -rf "$TEST_ROOT"
}