From 23b7df5fa97d0bfc5c3a6932aff50c1420318fa5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Andr=C3=A9=20Padez?= <andre.padez@gmail.com>
Date: Wed, 22 Jul 2026 15:05:23 +0000
Subject: [PATCH] task runner: kill the whole process tree on Stop

proc.kill() only signalled the direct child (bash), leaving a running ffmpeg
grandchild orphaned and still encoding. now killTree() walks /proc, SIGTERMs the
whole subtree (bash + ffmpeg), then SIGKILLs any straggler after 2s.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 src/servers/api/tasks/task-executor.ts | 56 +++++++++++++++++++++++++-
 1 file changed, 54 insertions(+), 2 deletions(-)

diff --git a/src/servers/api/tasks/task-executor.ts b/src/servers/api/tasks/task-executor.ts
index 589a30c2..ed951787 100644
--- a/src/servers/api/tasks/task-executor.ts
+++ b/src/servers/api/tasks/task-executor.ts
@@ -1,6 +1,6 @@
 import type { ServerWebSocket } from 'bun';
 import { join, isAbsolute } from 'node:path';
-import { mkdirSync, writeFileSync, chmodSync, rmSync } from 'node:fs';
+import { mkdirSync, writeFileSync, chmodSync, rmSync, readdirSync, readFileSync } from 'node:fs';
 import { getTaskByDirName } from './task-files';
 import { getHomeDirForRole, DATA_PATH } from '../../data-path';
 import { buildSandboxPrefix, buildRunuserSuffix } from '../../sidecar/sandbox';
@@ -36,6 +36,58 @@ type OutMessage =
 // Active processes per WebSocket
 const activeProcs = new WeakMap<ServerWebSocket<WSData>, { proc: ReturnType<typeof Bun.spawn>; kill: () => void }>();
 
+// Every descendant PID of `root`, from a single /proc walk (Linux). Bun's proc.kill() signals only
+// the direct child (bash), so a running grandchild like ffmpeg is left orphaned and keeps going.
+function descendantPids(root: number): number[] {
+  const childrenOf = new Map<number, number[]>();
+  let entries: string[];
+  try {
+    entries = readdirSync('/proc');
+  } catch {
+    return [];
+  }
+  for (const entry of entries) {
+    const pid = Number(entry);
+    if (!Number.isInteger(pid)) continue;
+    try {
+      const stat = readFileSync(`/proc/${pid}/stat`, 'utf8');
+      // "pid (comm) state ppid …" — comm may contain spaces/parens, so parse after the last ')'.
+      const fields = stat.slice(stat.lastIndexOf(')') + 2).split(' ');
+      const ppid = Number(fields[1]);
+      if (!Number.isInteger(ppid)) continue;
+      const list = childrenOf.get(ppid);
+      if (list) list.push(pid);
+      else childrenOf.set(ppid, [pid]);
+    } catch {
+      // process vanished mid-scan
+    }
+  }
+  const out: number[] = [];
+  const stack = [root];
+  while (stack.length > 0) {
+    const parent = stack.pop()!;
+    for (const child of childrenOf.get(parent) ?? []) {
+      out.push(child);
+      stack.push(child);
+    }
+  }
+  return out;
+}
+
+// Terminate a process and its whole subtree: SIGTERM everything, then SIGKILL stragglers after a
+// grace period. Killing the root (bash) too stops it from spawning the next file mid-batch.
+function killTree(root: number) {
+  const pids = [root, ...descendantPids(root)];
+  for (const pid of pids) {
+    try { process.kill(pid, 'SIGTERM'); } catch { /* already gone */ }
+  }
+  setTimeout(() => {
+    for (const pid of pids) {
+      try { process.kill(pid, 'SIGKILL'); } catch { /* gone */ }
+    }
+  }, 2000);
+}
+
 import { tmpdir } from 'node:os';
 
 function send(ws: ServerWebSocket<WSData>, msg: OutMessage) {
@@ -179,7 +231,7 @@ async function handleRun(ws: ServerWebSocket<WSData>, msg: RunMessage) {
     activeProcs.set(ws, {
       proc,
       kill: () => {
-        try { proc.kill(); } catch { /* already dead */ }
+        try { killTree(proc.pid); } catch { /* already dead */ }
       },
     });