Cannot run in multiple worker theads

[Bahlinc-Dev]

Is this package context aware?
It does run in node worker threads, but only a single thread.

I have an appsink in my pipeline, if I run 2 or more threads it will segfault during init:

Thread 1: pipeline.findChild('sink')
Result: OK

Thread 2: pipeline.findChild('sink')
Result: FATAL ERROR: v8::HandleScope::CreateHandle() Cannot create a handle without a HandleScope

[Bahlinc-Dev]

I know why you didn't see this before, it's tricky to reproduce.

I was trying to put together a simple test case and found it didn't happen all time, where as on our production pipeline it fails immediately. Then I realized it depends on what elements you use, how many instances, what you interact with on the pipeline, etc, but it will fail eventually at some point if you are using multiple instances across threads.

Here is the simplest test case I could produce that DOES fail, but you may have to run it a few times.
It creates 2 threads, and 2 instances per thread.

To run:

1. put index.js, worker.js, and the samplevideo from ./examples/loop/samplevideo.mp4 in a folder
2. start with UV_THREADPOOL_SIZE=10 node ./index.js

// index.js

const { Worker } = require('worker_threads');
async function startWorker(script, id) {
    return new Promise((resolve, reject) => {
        const worker = new Worker(script, { workerData: { id: id } });
        worker.on('online', () => {
            console.log(`Worker ${worker.threadId} online`);
            resolve(worker);
        });
        worker.on('message', msg => {
            console.log('Worker msg: ' + msg);
        });
    });
}

async function run() {
    const makeWorkers = 2
    for (let i = 0; i < makeWorkers; i += 1) {
        const worker = await startWorker('./worker.js', i);
    }
}
run();

// worker.js

const gstreamer = require('gstreamer-superficial');
const { parentPort, workerData } = require('worker_threads');
function createInstance(instanceId) {
    parentPort.postMessage(`CREATE INSTANCE Worker id:${workerData.id} instanceId:${instanceId}`);
    // simple broken pipeline
    let pipeline = new gstreamer.Pipeline(`
        filesrc location=./samplevideo.mp4 
        ! qtdemux 
        ! mp4mux
        ! appsink name=sink
    `);
    pipeline.play();
    const sink = pipeline.findChild('sink');
    let bufCount = 0;
    function onMp4Data(buf, caps) {
        if (buf) {
            bufCount += 1;
            if (!(bufCount % 50)) {
                parentPort.postMessage(`got 50 buffers in worker id:${workerData.id} cnt:${bufCount}`);
            }
            sink.pull(onMp4Data);
        }
    }
    sink.pull(onMp4Data);
    pipeline.pollBus( (msg) => {
        switch (msg.type) {
            case 'eos':
                console.log('End of Stream, making another instance');
                    createInstance(instanceId);
                break;
        }
    });
}
createInstance(1);
createInstance(2);

The full error on my machine is:

Worker 1 online
Worker msg: CREATE INSTANCE Worker id:0 instanceId:1
Worker msg: CREATE INSTANCE Worker id:0 instanceId:2
Worker 2 online
Worker msg: CREATE INSTANCE Worker id:1 instanceId:1
Worker msg: CREATE INSTANCE Worker id:1 instanceId:2
Worker msg: got 50 buffers in worker id:0 cnt:100
Worker msg: got 50 buffers in worker id:0 cnt:100
Worker msg: got 50 buffers in worker id:1 cnt:100
Worker msg: got 50 buffers in worker id:1 cnt:100
Worker msg: got 50 buffers in worker id:0 cnt:200
Worker msg: got 50 buffers in worker id:0 cnt:200
Worker msg: got 50 buffers in worker id:1 cnt:200
Worker msg: got 50 buffers in worker id:1 cnt:200
Worker msg: got 50 buffers in worker id:0 cnt:300
Worker msg: got 50 buffers in worker id:0 cnt:300
Worker msg: got 50 buffers in worker id:1 cnt:300
Worker msg: got 50 buffers in worker id:1 cnt:300
FATAL ERROR: v8::HandleScope::CreateHandle() Cannot create a handle without a HandleScope
End of Stream, making another instance
 1: 0xb00e10 node::Abort() [node]
Worker msg: CREATE INSTANCE Worker id:0 instanceId:1
 2: 0xa1823b node::FatalError(char const*, char const*) [node]
 3: 0xceddda v8::Utils::ReportApiFailure(char const*, char const*) [node]
 4: 0xe590a2 v8::internal::HandleScope::Extend(v8::internal::Isolate*) [node]
 5: 0xcef491 v8::HandleScope::CreateHandle(v8::internal::Isolate*, unsigned long) [node]
 6: 0x7f1d84690314  [/home/rc/iftest/node_modules/gstreamer-superficial/build/Release/gstreamer-superficial.node]
 7: 0xd4b6de  [node]
 8: 0xd4beb3 v8::internal::Builtins::InvokeApiFunction(v8::internal::Isolate*, bool, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::Object>, int, v8::internal::Handle<v8::internal::Object>*, v8::internal::Handle<v8::internal::HeapObject>) [node]
 9: 0xe1a0ab  [node]
10: 0xe1a9ae v8::internal::Execution::New(v8::internal::Isolate*, v8::internal::Handle<v8::internal::Object>, v8::internal::Handle<v8::internal::Object>, int, v8::internal::Handle<v8::internal::Object>*) [node]
11: 0xd0a27c v8::Function::NewInstanceWithSideEffectType(v8::Local<v8::Context>, int, v8::Local<v8::Value>*, v8::SideEffectType) const [node]
12: 0x7f1d8469140d GObjectWrap::NewInstance(Nan::FunctionCallbackInfo<v8::Value> const&, _GObject*) [/home/rc/iftest/node_modules/gstreamer-superficial/build/Release/gstreamer-superficial.node]
13: 0x7f1d84696b7b Pipeline::FindChild(Nan::FunctionCallbackInfo<v8::Value> const&) [/home/rc/iftest/node_modules/gstreamer-superficial/build/Release/gstreamer-superficial.node]
14: 0x7f1d84692a5c  [/home/rc/iftest/node_modules/gstreamer-superficial/build/Release/gstreamer-superficial.node]
15: 0xd4a82e  [node]
16: 0xd4bc4f v8::internal::Builtin_HandleApiCall(int, unsigned long*, v8::internal::Isolate*) [node]
17: 0x15e7dd9  [node]
Aborted (core dumped)

[Bahlinc-Dev]

Note that you can create as many instances as you want within a SINGLE thread, and it will work fine. This only happens when you have instances in different threads.

[dturing]

@Bahlinc-Dev node-gstreamer-superficial should be context-aware since #50 was closed- but i can't really test. do you have a suggestion what to do about this (or ideally a PR)? (i have little time to dig into this at all these days).

[onfire4g05]

I also am having this issue when using workers using the latest version.