highascg

Canonical `run.sh` (repo)

Use tools/casparcg-run.sh in this repository — it replaces the older pattern (RET=5 + tight relaunch) with:

</dev/null on the Caspar binary so stdin never blocks odd shutdown paths.
CASPAR_RESTART_GRACE_SEC (default 2) between exit and the next start so AMCP 5250 / CEF can release (tune to 3–5 if the second start fails with “address already in use”).
CASPAR_RESTART_EXIT_CODES (default 5) — space-separated list if your build uses another restart status.

Install on the playout host:

sudo install -m 0755 tools/casparcg-run.sh /opt/casparcg/run.sh

If AMCP RESTART still “hangs” (no second process, log stops): the Caspar process is stuck inside teardown — run.sh blocks on the first casparcg. pkill -f 'bin/casparcg.*config/casparcg.config' (main process only) so CASPAR_RESPAWN=1 can start again. Also try: close HTML producers before restart, wipe cef-cache, shorten command-line-args / disable GPU in config, or use KILL instead of RESTART.

If RESTART exits with a code other than 5 (wrapper exits and does not relaunch): run once, send RESTART, then echo $? in a wrapper, or check logs — set e.g. export CASPAR_RESTART_EXIT_CODES="5 42" before run.sh.

Example Openbox autostart (reference)

casparcg@serwer:~$ cat /home/casparcg/.config/openbox/autostart
#!/bin/bash

export DISPLAY=:0
export XAUTHORITY=/home/casparcg/.Xauthority

# Disable screen blanking
xset s off
xset s noblank
xset -dpms

# Hide mouse cursor
unclutter -idle 1 -root &

# --- Single instance lock (prevents duplicate starts) ---
(
  exec 9>/tmp/caspar-openbox-autostart.lock
  if ! flock -n 9; then
    exit 0
  fi

  ulimit -c unlimited

  cd /opt/casparcg || exit 1

  /usr/bin/casparcg-scanner &

  # Clean cache before start
  rm -rf /opt/casparcg/cef-cache/* 2>/dev/null

  # --- START CASPARCG (set CASPAR_RESPAWN=1 while debugging CEF crashes) ---
  # export CASPAR_RESPAWN=1
  # export CASPAR_RESTART_SLEEP=5
  bash /opt/casparcg/run.sh >> /tmp/caspar.log 2>&1

) &

Optional before run.sh: export CASPAR_RESTART_GRACE_SEC=3 if the child often fails to bind 5250 on the immediate relaunch.

Verify the loader (on the server)

After export LD_LIBRARY_PATH=/opt/casparcg/lib the same way as run.sh:

ldd /opt/casparcg/bin/casparcg | head -40
ldd /opt/casparcg/lib/libcef.so | egrep 'x264|ffmpeg|icu|nss|glib'

Anything you want only from your tree (e.g. libx264) must resolve to a path under /opt/casparcg/lib. If you still see /usr/lib/.../libx264.so, copy or symlink your build’s libx264.so* into /opt/casparcg/lib with matching SONAME (e.g. libx264.so.164), or rebuild Caspar/FFmpeg so dependencies are bundled there.

Check the binary’s runpath (should prefer your lib dir if you linked that way):

readelf -d /opt/casparcg/bin/casparcg | egrep 'RPATH|RUNPATH'

About the `partition_root.h` / `libx264` crash

Stacks often show both libcef.so and /usr/lib/.../libx264.so — that mix can trigger hard-to-debug allocator / alignment failures inside Chromium’s PartitionAlloc. Pinning libx264 (and any other codec libs Caspar loads) under /opt/casparcg/lib with a strict LD_LIBRARY_PATH is the first thing to try; if it still crashes, rebuild/run with a single consistent toolchain (same glibc, same build flags) for CEF, Caspar, and FFmpeg.

Segfault right after “Using CEF cache path”

Often still GPU / driver / bad cache. With CASPAR_RESPAWN=1, try between crashes: wipe /opt/casparcg/cef-cache/*, set <enable-gpu>false</enable-gpu> under <html> in casparcg.config (or trim <command-line-args>), confirm LD_LIBRARY_PATH only sees your libEGL.so / libGLESv2.so next to libcef.so. Use cores under ulimit -c unlimited with the autostart snippet above if you need a backtrace.

Autostart log line

Use a single append redirect for the log (avoid duplicate > / 2>&1), as in the autostart snippet above.

Restart Caspar after a config change

The run.sh loop runs Caspar again when the process exits with a restart status (default 5; override with CASPAR_RESTART_EXIT_CODES). tools/casparcg-run.sh adds a short CASPAR_RESTART_GRACE_SEC pause before the next start. Saving a new casparcg.config on disk does not by itself exit the process, so Openbox will not re-run run.sh until you log in again.

Preferred (no extra process kill): send AMCP RESTART after the file is written — Caspar reloads/restarts according to your build. HighAsCG Settings → System → Write & restart does exactly that (writes XML, then RESTART on port 5250). Or manually, from any host that reaches the playout machine:

# Replace host if Caspar AMCP is not local
printf 'RESTART\r\n' | nc -N 127.0.0.1 5250

After a segfault while tuning CEF / config: enable CASPAR_RESPAWN=1 in autostart (see commented export lines above). Then each crash waits CASPAR_RESTART_SLEEP seconds and starts Caspar again — new casparcg.config on disk is read on every start. To retry immediately without waiting for a crash: pkill -f '/opt/casparcg/bin/casparcg' (only the casparcg binary; leave the bash run.sh parent running).

Hard restart via AMCP: RESTART on port 5250 (see above) — useful when Caspar stays up and you only changed config.

If you edited run.sh itself (e.g. LD_LIBRARY_PATH): the running bash already loaded the old script — pkill casparcg is not enough. Kill the run.sh bash (e.g. find its PID, kill <pid>), then start again from a terminal with DISPLAY=:0 or re-login.

Scanner: restarting Caspar does not restart casparcg-scanner; restart that separately if you change scanner config (pkill casparcg-scanner then start it again, or log out/in).

This site is open source. Improve this page.