| 22 Jan 2026 |
 K900 | I think RT kernels may kill processes that eat too much CPU time but I don't think this should ever trip on normal kernels | 09:16:50 |
 Marie | nothing in dmesg | 09:18:20 |
| Marie | but i just saw this in the journal /nix/store/slfhbw60dvv03rlcmr5r2kzhlnrbaxmv-xdg-desktop-portal-1.20.3/libexec/xdg-desktop-portal[2091]: Realtime error: Could not map tid 23 of pid 2: Process ids could not be found: 23, not sure if thats because it dies | 09:18:52 |
| K900 | That shouldn't make it sigkill itself though | 09:19:13 |
| K900 | Have you looked at the actual scripts inside at all? | 09:20:22 |
| Marie | you mean inside the flatpak? | 09:21:31 |
| Marie | it's just this wrapper script which handles self updating
#!/bin/sh
LAUNCHER_DIR="$XDG_DATA_HOME/Hytale"
LAUNCHER_BIN="$LAUNCHER_DIR/hytale-launcher"
BUNDLED_HASH_FILE="$LAUNCHER_DIR/.bundled_hash"
mkdir -p "$LAUNCHER_DIR"
# Compute hash of bundled binary to detect Flatpak package updates
BUNDLED_HASH=$(sha256sum /app/bin/hytale-launcher | cut -d" " -f1)
# Copy bundled binary if: no binary exists, OR the Flatpak package was updated
# This preserves self-updates while still applying Flatpak package updates
if [ ! -x "$LAUNCHER_BIN" ] || [ ! -f "$BUNDLED_HASH_FILE" ] || [ "$(cat "$BUNDLED_HASH_FILE")" != "$BUNDLED_HASH" ]; then
install -m755 /app/bin/hytale-launcher "$LAUNCHER_BIN"
echo "$BUNDLED_HASH" > "$BUNDLED_HASH_FILE"
fi
exec "$LAUNCHER_BIN" "$@"
| 09:23:31 |
| K900 | And this is the process that gets sigkilled? | 09:23:54 |
| K900 | Hmm I wonder if killsnoop is actually the weird one here and doesn't catch the exec | 09:24:20 |
| Marie | the script should work fine, the launcher window pops up for a split second | 09:24:38 |
| K900 | Would be nice to maybe get a stack trace from when it actually dies | 09:25:10 |
| Marie | i tried that with bpftrace, but it only gave me pointers | 09:25:47 |
| Marie | it dies inside the launchers code, so no debug info | 09:25:57 |
