ci: Consolidate build-host script

.gitlab/ci/build-host.sh:
Change all script-local variables to lower-case and make some of them overridable using environment variables (by
providing defaults).
Break down overly long commands by splitting them into a list of strings.
Use local variables where possible.
Change `main()` to use rsync instead of cp to copy the project to the build location more generically.
Change `main()` to use rsync instead of cp to copy the build artifacts on the VM from the project's build directory to
the output.
Remove all unnecessary `function` keywords for function declarations.
Replace the dependency on libisoburn's `xorriso` with libarchive's `bsdtar` and util-linux's `blkid` in
`prepare_boot()`.
Add `print_section_start()` and `print_section_end()` to reduce code duplication and error potential when printing lines
for gitlab's collapsible sections (https://docs.gitlab.com/ee/ci/jobs/#custom-collapsible-sections).
Document the script's behavior and expectations.
Document the understood environment variables and add links to documentation on understood units (in case of size
units).

.gitlab/ci/build-host.sh

@@ -1,88 +1,181 @@
 #!/usr/bin/env bash
-# build-host.sh runs build-inside-vm.sh in a qemu VM running the latest Arch installer iso
 #
-# nounset: "Treat unset variables and parameters [...] as an error when performing parameter expansion."
+# This script runs a build script in a QEMU VM using the latest Arch Linux installation medium.
-# errexit: "Exit immediately if [...] command exits with a non-zero status."
+# The build script is expected to create an './output' directory in the project's directory (when running in the VM) and
-set -o nounset -o errexit
+# place any build artifacts there.
-readonly MIRROR="https://mirror.pkgbuild.com"
+# After the build script has finished this script will copy all artifacts to a (local) './output' directory and shutdown
+# the VM.
+#
+# Dependencies:
+# - coreutils
+# - curl
+# - libarchive
+# - qemu-headless
+# - util-linux
+#
+# Considered environment variables:
+# ARCHISO_COW_SPACE_SIZE: The amount of RAM to allocate for the copy-on-write space used by archiso (defaults to 1g -
+# see https://man.archlinux.org/man/tmpfs.5 for understood units)
+# ARCHITECTURE: A string to set the CPU architecture (defaults to x86_64)
+# BUILD_SCRIPT: A script that will be called on the host (defaults to ./build-inside-vm.sh)
+# BUILD_SCRIPT_ARGS: The arguments to BUILD_SCRIPT (as a space delimited list)
+# PACKAGE_LIST: A space delimited list of packages to install to the virtual machine
+# PACMAN_MIRROR: The pacman mirror to use (defaults to "https://mirror.pkgbuild.com")
+# QEMU_DISK_SIZE: A string given to fallocate to create a scratch disk to build in (defaults to 8G - see
+# https://man.archlinux.org/man/fallocate.1 for understood units)
+# QEMU_VM_MEMORY: The amount of RAM (in MiB) allocated for the QEMU virtual machine (defaults to 1024)
+# QEMU_LOGIN_TIMEOUT: The maximum time (in seconds) to wait for the initial prompt in the VM to appear (defaults to 60)
+# QEMU_PACKAGES_TIMEOUT: The maximum time (in seconds) to wait for output from pacman when installing packages (defaults
+# to 120)
+# QEMU_BUILD_TIMEOUT: The maximum time (in seconds) to wait for output from the build script (defaults to 1800)
+# QEMU_COPY_ARTIFACTS_TIMEOUT: The maximum time (in seconds) to wait for output from the action of copying the build
+# artifacts from the VM to a local directory (defaults to 60)
 
+
+set -euo pipefail
+
+readonly orig_pwd="${PWD}"
+readonly output="${PWD}/output"
+
+# variables with presets/ environmental overrides
+arch="${ARCHITECTURE:-x86_64}"
+script="${BUILD_SCRIPT:-./build-inside-vm.sh}"
+script_args="${BUILD_SCRIPT_ARGS:-}"
+mirror="${PACMAN_MIRROR:-https://mirror.pkgbuild.com}"
+disk_size="${QEMU_DISK_SIZE:-8G}"
+vm_memory="${QEMU_VM_MEMORY:-1024}"
+login_timeout="${QEMU_LOGIN_TIMEOUT:-60}"
+packages_timeout="${QEMU_PACKAGES_TIMEOUT:-120}"
+build_timeout="${QEMU_BUILD_TIMEOUT:-1800}"
+copy_artifacts_timeout="${QEMU_COPY_ARTIFACTS_TIMEOUT:-60}"
+cow_space_size="${ARCHISO_COW_SPACE_SIZE:-1g}"
+packages="${PACKAGE_LIST:-}"
+
+# variables without presets/ environmental overrides
+iso=""
+iso_volume_id=""
 tmpdir=""
-
-function init() {
-  readonly ORIG_PWD="${PWD}"
-  readonly OUTPUT="${PWD}/output"
 tmpdir="$(mktemp --dry-run --directory --tmpdir="${PWD}/tmp")"
-  mkdir -p "${OUTPUT}" "${tmpdir}"
 
+print_section_start() {
+  # gitlab collapsible sections start: https://docs.gitlab.com/ee/ci/jobs/#custom-collapsible-sections
+  local _section _title
+  _section="${1}"
+  _title="${2}"
+
+  printf "\e[0Ksection_start:%(%s)T:%s\r\e[0K%s\n" '-1' "${_section}" "${_title}"
+}
+
+print_section_end() {
+  # gitlab collapsible sections end: https://docs.gitlab.com/ee/ci/jobs/#custom-collapsible-sections
+  local _section
+  _section="${1}"
+
+  printf "\e[0Ksection_end:%(%s)T:%s\r\e[0K\n" '-1' "${_section}"
+}
+
+init() {
+  print_section_start "create_dirs" "Create required directories"
+
+  mkdir -p "${output}" "${tmpdir}"
   cd "${tmpdir}"
+
+  print_section_end "create_dirs"
 }
 
 # Do some cleanup when the script exits
-function cleanup() {
+cleanup() {
+  print_section_start "cleanup" "Cleaning up"
+
   rm -rf -- "${tmpdir}"
   jobs -p | xargs --no-run-if-empty kill
+
+  print_section_end "cleanup"
 }
 trap cleanup EXIT
 
 # Use local Arch iso or download the latest iso and extract the relevant files
-function prepare_boot() {
+prepare_boot() {
-  local iso
+  local _latest_iso _iso
-  local isos=()
+  local _isos=()
+
+  print_section_start "prepare_boot" "Prepare boot media"
 
   # retrieve any local images and sort them
-  for iso in "${ORIG_PWD}/"archlinux-*-x86_64.iso; do
+  for _iso in "${orig_pwd}/"archlinux-*-"${arch}.iso"; do
-    if [[ -f "$iso" ]]; then
+    if [[ -f "${_iso}" ]]; then
-      isos+=("${iso}")
+      _isos+=("${_iso}")
     fi
   done
-  if (( ${#isos[@]} >= 1 )); then
+  if (( ${#_isos[@]} >= 1 )); then
-    ISO="$(printf '%s\n' "${isos[@]}" | sort -r | head -n1)"
+    iso="$(printf '%s\n' "${_isos[@]}" | sort -r | head -n1)"
-    printf "Using local iso: %s\n" "$ISO"
+    printf "Using local iso: %s\n" "$iso"
   fi
 
-  if (( ${#isos[@]} < 1 )); then
+  if (( ${#_isos[@]} < 1 )); then
-    LATEST_ISO="$(curl -fs "${MIRROR}/iso/latest/" | grep -Eo 'archlinux-[0-9]{4}\.[0-9]{2}\.[0-9]{2}-x86_64.iso' | head -n 1)"
+    _latest_iso="$(
-    if [[ -z "${LATEST_ISO}" ]]; then
+        curl -fs "${mirror}/iso/latest/" | \
-      echo "Error: Couldn't find latest iso'"
+        grep -Eo "archlinux-[0-9]{4}\.[0-9]{2}\.[0-9]{2}-${arch}.iso" | \
+        head -n 1
+    )"
+    if [[ -z "${_latest_iso}" ]]; then
+      echo "Error: Could not find latest iso"
       exit 1
     fi
-    curl -fO "${MIRROR}/iso/latest/${LATEST_ISO}"
+    curl -fO "${mirror}/iso/latest/${_latest_iso}"
-    ISO="${PWD}/${LATEST_ISO}"
+    iso="${PWD}/${_latest_iso}"
   fi
 
-  # We need to extract the kernel and initrd so we can set a custom cmdline:
+  # Extract the kernel and initrd so that a custom kernel cmdline can be set:
-  # console=ttyS0, so the kernel and systemd sends output to the serial.
+  # console=ttyS0, so that the kernel and systemd send output to the serial.
-  xorriso -osirrox on -indev "${ISO}" -extract arch/boot/x86_64 .
+  bsdtar -x -f "${iso}" -C . "arch/boot/${arch}"
-  ISO_VOLUME_ID="$(xorriso -indev "${ISO}" |& awk -F : '$1 ~ "Volume id" {print $2}' | tr -d "' ")"
+  iso_volume_id="$(blkid -s LABEL -o value "${iso}")"
+
+  print_section_end "prepare_boot"
 }
 
-function start_qemu() {
+start_qemu() {
+  local _kernel_params=(
+    "archisobasedir=arch"
+    "archisolabel=${iso_volume_id}"
+    "cow_spacesize=${cow_space_size}"
+    "ip=dhcp"
+    "net.ifnames=0"
+    "console=ttyS0"
+    "mirror=${mirror}"
+  )
+
+  print_section_start "start_qemu" "Start VM using QEMU"
+
   # Used to communicate with qemu
   mkfifo guest.out guest.in
   # We could use a sparse file but we want to fail early
-  fallocate -l 8G scratch-disk.img
+  fallocate -l "${disk_size}" scratch-disk.img
 
   { qemu-system-x86_64 \
     -machine accel=kvm:tcg \
     -smp "$(nproc)" \
-    -m 4096 \
+    -m "${vm_memory}" \
-    -device virtio-net-pci,romfile=,netdev=net0 -netdev user,id=net0 \
+    -device virtio-net-pci,romfile=,netdev=net0 \
-    -kernel vmlinuz-linux \
+    -netdev user,id=net0 \
-    -initrd initramfs-linux.img \
+    -kernel "arch/boot/${arch}/vmlinuz-linux" \
-    -append "archisobasedir=arch archisolabel=${ISO_VOLUME_ID} cow_spacesize=4G ip=dhcp net.ifnames=0 console=ttyS0 mirror=${MIRROR}" \
+    -initrd "arch/boot/${arch}/initramfs-linux.img" \
+    -append "${_kernel_params[*]}" \
     -drive file=scratch-disk.img,format=raw,if=virtio \
-    -drive file="${ISO}",format=raw,if=virtio,media=cdrom,read-only=on \
+    -drive "file=${iso},format=raw,if=virtio,media=cdrom,read-only=on" \
-    -virtfs "local,path=${ORIG_PWD},mount_tag=host,security_model=none" \
+    -virtfs "local,path=${orig_pwd},mount_tag=host,security_model=none" \
     -monitor none \
     -serial pipe:guest \
     -nographic || kill "${$}"; } &
 
   # We want to send the output to both stdout (fd1) and fd10 (used by the expect function)
   exec 3>&1 10< <(tee /dev/fd/3 <guest.out)
+
+  print_section_end "start_qemu"
 }
 
 # Wait for a specific string from qemu
-function expect() {
+expect() {
   local length="${#1}"
   local i=0
   local timeout="${2:-30}"
@@ -103,17 +196,28 @@ function expect() {
 }
 
 # Send string to qemu
-function send() {
+send() {
   echo -en "${1}" >guest.in
 }
 
-function main() {
+main() {
+  local _pacman_command=(
+    "pacman -Fy &&"
+    "pacman -Syu --ignore"
+    "\$(pacman -Fq --machinereadable /usr/lib/modules/"
+    "| awk 'BEGIN { FS =\"\\\0\";ORS=\",\" }; { print \$2 }'"
+    "| sort -ut , | head -c -2)"
+    "--noconfirm --needed ${packages}\n"
+  )
+
   init
   prepare_boot
   start_qemu
 
+  print_section_start "init_build_environment" "Initialize build environment"
+
   # Login
-  expect "archiso login:" 60
+  expect "archiso login:" "${login_timeout}"
   send "root\n"
   expect "# "
 
@@ -128,7 +232,7 @@ function main() {
   expect "# "
   send "mkfs.ext4 /dev/vda && mkdir /mnt/scratch-disk/ && mount /dev/vda /mnt/scratch-disk && cd /mnt/scratch-disk\n"
   expect "# "
-  send "cp -a -- /mnt/project/{.gitlab,archiso,configs,scripts} .\n"
+  send "rsync -a --exclude tmp --exclude .git -- /mnt/project/ .\n"
   expect "# "
   send "mkdir pkg && mount --bind pkg /var/cache/pacman/pkg\n"
   expect "# "
@@ -138,22 +242,38 @@ function main() {
   expect "# "
 
   # Explicitly lookup mirror address as we'd get random failures otherwise during pacman
-  send "curl -sSo /dev/null ${MIRROR}\n"
+  send "curl -sSo /dev/null ${mirror}\n"
   expect "# "
 
+  print_section_end "init_build_environment"
+  print_section_start "install_packages" "Install packages"
+
+  if [[ -n "${packages}" ]]; then
     # Install required packages
-  send "pacman -Fy && pacman -Syu --ignore \$(pacman -Fq --machinereadable /usr/lib/modules/ | awk 'BEGIN { FS = \"\\\0\";ORS=\",\" }; { print \$2 } ' | sort -ut , | head -c -2) --noconfirm --needed qemu-headless jq dosfstools erofs-utils e2fsprogs libisoburn mtools squashfs-tools zsync\n"
+    send "${_pacman_command[*]}"
-  expect "# " 120
+    expect "# " "${packages_timeout}"
+  fi
+
+  print_section_end "install_packages"
 
   ## Start build and copy output to local disk
-  send "bash -x ./.gitlab/ci/build-inside-vm.sh ${PROFILE} ${BUILDMODE}\n "
+  send "bash -x ${script} ${script_args}\n "
-  expect "# " 2400 # mksquashfs can take a very long time
+  expect "# " "${build_timeout}"
-  send "cp -r --preserve=mode,timestamps -- output /mnt/project/tmp/$(basename "${tmpdir}")/\n"
+
-  expect "# " 60
+  print_section_start "move_artifacts" "Move artifacts to output directory"
-  mv output/* "${OUTPUT}/"
+
+  send "rsync -av -- output /mnt/project/tmp/$(basename "${tmpdir}")/\n"
+  expect "# " "${copy_artifacts_timeout}"
+  mv -- output/* "${output}/"
+
+  print_section_end "move_artifacts"
+  print_section_start "shutdown" "Shutdown the VM"
 
   # Shutdown the VM
   send "systemctl poweroff -i\n"
   wait
+
+  print_section_end "shutdown"
 }
+
 main