#!/usr/bin/env bash # # Build and flash the whole ClockClock 24 wall, one board at a time. # macOS and Linux. Written for bash 3.2, which is what macOS still ships. # # Eight boards, and only one of them has OTA. The other seven go over USB, which # means moving the lead between each - so this compiles everything first, then # walks you through the uploads with a pause at each board. # # Two things it does deliberately: # # COMPILE EVERYTHING FIRST. A compile error found halfway through flashing # leaves you with a wall running two firmwares and a board in your hand. All # eight are built before anything is uploaded. # # THE ORDER YOU ASK FOR IS THE ORDER IT USES. a,b,c,d,e,f,g,h by default - # the order the boards sit on the wall, which is the order you have them in # your hand. # # When the order DOES matter: updating a wall that stays powered while you work # through it. The mode is an integer on the sync bus and new modes are appended, # so a NEWER master can broadcast a mode an OLDER listener does not know, and # that listener silently drops the mode field and holds its last animation. The # other way round is harmless. If that is what you are doing, put the master # last: `-b a,b,c,e,f,g,h,d`. Flashing everything in one go with the wall down - # or with the modules out of their carriers - has no such window. # # Usage: # ./flash-all.sh # all eight, over USB, in order # ./flash-all.sh -b a,b,c # just these # ./flash-all.sh -b a,b,c,e,f,g,h,d # listeners first, master last # ./flash-all.sh -p /dev/ttyACM0 # skip the port prompt # ./flash-all.sh --build-only # check a change compiles for all eight # ./flash-all.sh --skip-build # upload what is already built # ./flash-all.sh -m cc24-board-d.local # master over the network, not USB # ./flash-all.sh -m 192.168.1.42 # ...by IP instead of mDNS set -euo pipefail # `readlink -f` is GNU-only; BSD readlink on macOS has no -f. dirname of the # invocation path is enough for a script that lives beside its configs. cd "$(dirname "${BASH_SOURCE[0]}")" BOARDS="a,b,c,d,e,f,g,h" # Empty = the master goes over USB like everything else, which is what you want # when the modules are out of the wall in front of you. -m opts into OTA. MASTER_HOST="" PORT="" BUILD_ONLY=0 SKIP_BUILD=0 while [ $# -gt 0 ]; do case "$1" in -b|--boards) BOARDS="$2"; shift 2 ;; -p|--port) PORT="$2"; shift 2 ;; -m|--master-host) MASTER_HOST="$2"; shift 2 ;; --build-only) BUILD_ONLY=1; shift ;; --skip-build) SKIP_BUILD=1; shift ;; -h|--help) sed -n '2,30p' "$0" | sed 's/^# \?//'; exit 0 ;; *) echo "Unknown option: $1 (try --help)" >&2; exit 2 ;; esac done # Board letter -> column and clock indices. Printed as we go, because a board # flashed into the wrong column is the one mistake that looks like a hardware # fault: the wall lights up, the time is simply wrong. board_info() { case "$1" in a) echo "column 0, clocks 0 / 2 / 4" ;; b) echo "column 1, clocks 1 / 3 / 5" ;; c) echo "column 2, clocks 6 / 8 / 10" ;; d) echo "column 3, clocks 7 / 9 / 11 [MASTER]" ;; e) echo "column 4, clocks 12 / 14 / 16" ;; f) echo "column 5, clocks 13 / 15 / 17" ;; g) echo "column 6, clocks 18 / 20 / 22" ;; h) echo "column 7, clocks 19 / 21 / 23" ;; *) echo "unknown board '$1'" >&2; return 1 ;; esac } # The wall, drawn as you are actually looking at it: from BEHIND, where the USB # sockets are - so board A is on the RIGHT. Columns are zero-based (A = 0), and # "column 4" read as the fourth one along lands on D, which is exactly the kind # of thing a picture settles and a number does not. BACK_ORDER="h g f e d c b a" col_of() { case "$1" in a) echo 0 ;; b) echo 1 ;; c) echo 2 ;; d) echo 3 ;; e) echo 4 ;; f) echo 5 ;; g) echo 6 ;; h) echo 7 ;; esac } wall_map() { local target="$1" col row L c cell line head pad k=0 tcol="" head=" " for L in $BACK_ORDER; do head="$head $(echo "$L" | tr '[:lower:]' '[:upper:]') " [ "$L" = "$target" ] && tcol=$k k=$((k + 1)) done printf '\n %sthe wall from BEHIND - the side you plug into%s\n\n' "$C_DIM" "$C_OFF" printf '%s\n' "$head" printf '%s\n' " ┌────┬────┬────┬────┬────┬────┬────┬────┐" for row in 0 1 2; do line=" row${row} │" for L in $BACK_ORDER; do col="$(col_of "$L")" c=$(( (col / 2) * 6 + row * 2 + (col % 2) )) if [ "$L" = "$target" ]; then # Reverse video on a terminal, *nn* when piped - the marker has to # survive being pasted into an issue. if [ -n "$C_REV" ]; then cell="$(printf '%s %2d %s' "$C_REV" "$c" "$C_OFF")" else cell="$(printf '*%2d*' "$c")"; fi else cell="$(printf ' %2d ' "$c")" fi line="${line}${cell}│" done printf '%s\n' "$line" done printf '%s\n' " └────┴────┴────┴────┴────┴────┴────┴────┘" if [ -n "$tcol" ]; then pad="$(printf '%*s' $((11 + tcol * 5)) '')" printf '%s└─ BOARD %s\n\n' "$pad" "$(echo "$target" | tr '[:lower:]' '[:upper:]')" fi } if [ -t 1 ]; then C_STEP=$'\e[36m'; C_OK=$'\e[32m'; C_WARN=$'\e[33m'; C_DIM=$'\e[90m'; C_OFF=$'\e[0m' C_REV=$'\e[7m' else C_STEP=""; C_OK=""; C_WARN=""; C_DIM=""; C_OFF="" C_REV="" fi step() { printf '\n%s=== %s%s\n' "$C_STEP" "$1" "$C_OFF"; } # mm:ss. Compiling eight boards is minutes, not seconds, and "has it hung?" is # the question these numbers exist to answer. fmt_dur() { printf '%dm%02ds' $(( $1 / 60 )) $(( $1 % 60 )); } ok() { printf '%s %s%s\n' "$C_OK" "$1" "$C_OFF"; } warn() { printf '%s %s%s\n' "$C_WARN" "$1" "$C_OFF"; } dim() { printf '%s %s%s\n' "$C_DIM" "$1" "$C_OFF"; } command -v esphome >/dev/null 2>&1 || { echo "esphome is not on PATH. Install it with: pip install esphome" >&2; exit 1; } ok "$(esphome version 2>&1 | head -1)" # No reordering. An order that quietly differs from the one you typed is worse # than one that is occasionally wrong - see the note at the top for when it # matters. IFS=',' read -r -a REQUESTED <<< "$BOARDS" ORDER=() WANT_MASTER=0 for b in "${REQUESTED[@]}"; do b="$(echo "$b" | tr -d '[:space:]' | tr '[:upper:]' '[:lower:]')" [ -z "$b" ] && continue board_info "$b" >/dev/null || exit 2 [ "$b" = "d" ] && WANT_MASTER=1 ORDER+=("$b") done [ ${#ORDER[@]} -eq 0 ] && { echo "No boards selected." >&2; exit 2; } # ------------------------------------------------------------------ build --- # Phase numbering adapts, so --skip-build does not announce a phase 1 that # never runs. PHASES=2 [ "$SKIP_BUILD" -eq 1 ] && PHASES=1 [ "$BUILD_ONLY" -eq 1 ] && PHASES=1 PHASE=0 RUN_START=$SECONDS COMPILE_SECS=0 UPLOAD_SECS=0 UPLOADS=0 if [ "$SKIP_BUILD" -eq 0 ]; then PHASE=$((PHASE + 1)) step "Phase $PHASE/$PHASES - compiling ${#ORDER[@]} board(s)" dim "Nothing is uploaded until they all build." n=0 phase_start=$SECONDS for b in "${ORDER[@]}"; do n=$((n + 1)) printf '\n [%d/%d] compiling board_%s.yaml (%s)\n' \ "$n" "${#ORDER[@]}" "$b" "$(board_info "$b")" t0=$SECONDS esphome compile "board_$b.yaml" ok "[$n/${#ORDER[@]}] board_$b built in $(fmt_dur $((SECONDS - t0)))" done COMPILE_SECS=$((SECONDS - phase_start)) ok "All ${#ORDER[@]} compiled in $(fmt_dur $COMPILE_SECS)." fi if [ "$BUILD_ONLY" -eq 1 ]; then step "Finished in $(fmt_dur $((SECONDS - RUN_START))) - --build-only, nothing uploaded" exit 0 fi # Pick a serial port: use the only one if there is only one, else ask. An empty # answer is fine - esphome will choose. pick_port() { [ -n "$PORT" ] && { echo "$PORT"; return; } local found=() # macOS calls them /dev/cu.*; Linux /dev/ttyUSB* or /dev/ttyACM*. Use cu. # rather than tty. on macOS: tty. blocks on open waiting for carrier detect, # which a USB-serial bridge never asserts. for dev in /dev/cu.usbmodem* /dev/cu.usbserial* /dev/cu.wchusbserial* \ /dev/cu.SLAB_USBtoUART* /dev/ttyUSB* /dev/ttyACM*; do [ -e "$dev" ] && found+=("$dev") done if [ ${#found[@]} -eq 1 ]; then warn "Using ${found[0]} (the only port present)" >&2 echo "${found[0]}"; return fi if [ ${#found[@]} -eq 0 ]; then warn "No serial ports found - is the board plugged in?" >&2 else warn "Ports: ${found[*]}" >&2 fi local answer read -r -p " Serial port (blank to let esphome choose): " answer