thread function don't excute

[Y0QIN]

I have a very strange problem by using pthread. After the thread is created, the thread function isn't executed. And this is an occasional problem that happend when I fresh or enter the page.
Has someone meet the same or similar issues?
Can someone give me some advice?

[sbc100]

Something this can happen if you don't yield on the main thread. In general threads can't start until the main thread yield (returns from to JS event loop).

If you don't want to break up you main thread code you can also use PROXY_TO_PTHREAD to PTHREAD_POOL_SIZE to solve this problem.

[Y0QIN]

@sbc100 Could you say something about why this happened.
I don't quite understand that the thread has used the join mode and will wait for the sub thread to finish. Why does this happen? Maybe it's because of the browser's event mechanism？

[sbc100]

What i'm describing might not be the problem you are seeing, but assuming it is, the problem is this:

In order to create a new thread the browser needs to create a new WebWorker and send messages to it. In order to do that the JS event loop needs to run, but the JS event loop only runs when you return all the way out of your program. So if you call pthread_create() (unless you use PROXY_TO_PTHREAD or PTHREAD_POOL_SIZE) the new thread won't actually start until you return out and let the JS event loop run.

@kripken is about to land change that will make this a lot more obvious: #9579

[sbc100]

If you enable PROXY_TO_PTHREAD and the problem goes away then you will know that you are seeing this problem.

[Y0QIN]

@sbc100 use PROXY_TO_PTHREAD to PTHREAD_POOL_SIZE means to -s PROXY_TO_PTHREAD or -s PTHREAD_POOL_SIZE=PROXY_TO_PTHREAD？
when I use -s PROXY_TO_PTHREAD -PTHREAD_POOL_SIZE=4 to re-compile my code , it success but it occur the following problem:
ffmpeg.worker.js:226 ReferenceError: wasmAlready is not defined
at Array.ASM_CONSTS (http://192.168.4.104/js/ffmpeg/ffmpeg.js?va08v:1:19399)
at _emscripten_asm_const_i (http://192.168.4.104/js/ffmpeg/ffmpeg.js?va08v:1:19880)
at wasm-function[1547]:0xb7d00
at Module._main (http://192.168.4.104/js/ffmpeg/ffmpeg.js?va08v:1:164816)
at ___call_main (http://192.168.4.104/js/ffmpeg/ffmpeg.js?va08v:1:31600)
at wasm-function[1447]:0xb0c56
at wasm-function[1426]:0xaf7a6
at Object.Module.dynCall_ii (http://192.168.4.104/js/ffmpeg/ffmpeg.js?va08v:1:166121)
at onmessage (http://192.168.4.104/js/ffmpeg/ffmpeg.worker.js:191:42)
onmessage @ ffmpeg.worker.js:226

that because I use EM_ASM in C code? I use EM_ASM to make sure my file has loaded already. How could I solve this problem?

int main()
{ 
    printf("########FFmpeg library import complete#########\n");
    EM_ASM(wasmAlready());
    memset(&g_context, 0, sizeof(struct _ffmpeg_context_t) * MAX_CHAN_NUM);
    emscripten_exit_with_live_runtime();
    return 0;
}

It seems like the main() is also run in the worker threads, so I can't use functions orvariable of the browser main (UI) thread.
I solved the error above.
But it still occur thread deadlock that the thread function do not excute after thread create ,before thread_jion

[sbc100]

Where are you defining wasmAlready? Since your code is now running in worker so you'd want to make sure wasmAlready is defined in the js file along with the module. One way to do include it via --pre-js or --post-js.

Alternatively forget about PROXY_TO_PTHREAD and just add PTHREAD_POOL_SIZE which should be simpler.

[Y0QIN]

@sbc100 I solved the error above about wasmAlready.
But it still occur thread deadlock that the thread function do not excute after thread create ,before thread_jion

[sbc100]

Did you try add PTHREAD_POOL_SIZE=N? ( where N > 1)

Can you run simple thread test? e.g. ./tests/runner.py browser.test_pthread_create?

[Y0QIN]

@sbc100 I have add PTHREAD_POOL_SIZE=4 already. And it test success when I run the simple test by cmd python ./tests/runner.py browser.test_pthread_create.
then I have try to use ALLOW_BLOCKING_ON_MAI,there is an error printed:
shared:ERROR: Assigning a non-existent settings attribute "ALLOW_BLOCKING_ON_MAI
N_THREAD"

In addition, my wasm files is normal in win7,but occasionally deadlock in win10

int FFMpegContextCreate(int iChannelId, int bYUVShow, int bOptimizeDecode, int codec_type,int iPlayId)
{
    if(iChannelId < 0 || iChannelId > MAX_CHAN_NUM) 
        return -1;
    int i;
    int ret;
    struct _ffmpeg_context_t* _context = &g_context[iChannelId];
    printf("start create\n");
    _context->bRun=1;
    _context->dwUser = _context;
    _context->hThreadAVPktDecode = ys_thread_create(2000, ThreadAVPktDecode, _context);
   ....
}
static void* ThreadAVPktDecode(void* dwUser){
    printf("do Decode thread start test\n");
    struct _ffmpeg_context_t* _context = (struct _ffmpeg_context_t *)dwUser;
    static int iTest = 0;
    printf("do Decode thread test\n");
    while (_context->bRun) {
         iTest ++;
         usleep(50000);
         printf("do Decode thread test %d\n",_context->bRun);
    }
    printf("do Decode thread end\n");
    return NULL;
}
int FFMpegContextDestroy(int iChannelId)
{
    printf("FFMpegContextDestroy begin ,iChannelId:%d \n",iChannelId);
    struct _ffmpeg_context_t* _context = &g_context[iChannelId];
    int i = 0;
    _context->bRun = 0;
    _context->nAvcodecDecode = 0;
    printf("FFMpegContextDestroy 000000111\n");
    if(_context->hThreadAVPktDecode!=NULL)
        ys_thread_join(_context->hThreadAVPktDecode);
    printf("FFMpegContextDestroy 000000222\n");
    ys_free(_context->hThreadAVPktDecode);
    _context->hThreadAVPktDecode = NULL;
    printf("FFMpegContextDestroy 1111\n");              
    ys_list_special_free_http_data(_context->pListAVPkt);
}

ys_thread_create and ys_thread_join are functions to handle pthread
And when it ocuur deadlock there isn't a log of do Decode thread start

[stale]

This issue has been automatically marked as stale because there has been no activity in the past year. It will be closed automatically if no further activity occurs in the next 30 days. Feel free to re-open at any time if this issue is still relevant.