Re: [PATCH resend 2/2] winegstreamer: Move broad callback handling code to gst_cbs.

[Zebediah Figura]

On 3/17/20 12:07 PM, Derek Lesho wrote:
> +BOOL is_wine_thread(void)
> +{
> - return pthread_getspecific(wine_gst_key) != NULL;
>
> +}
> -
>
> +pthread_mutex_t cb_list_lock = PTHREAD_MUTEX_INITIALIZER;
> +pthread_cond_t cb_list_cond = PTHREAD_COND_INITIALIZER;
> +struct list cb_list = LIST_INIT(cb_list);
> -
>
> +void CALLBACK perform_cb(TP_CALLBACK_INSTANCE *instance, void *user)
> +{
> - struct cb_data *cbdata = user;
> -
> - if (cbdata->type < GSTDEMUX_MAX)
> -
>    perform_cb_gstdemux(cbdata);
> -
> - pthread_mutex_lock(&cbdata->lock);
> - cbdata->finished = 1;
> - pthread_cond_broadcast(&cbdata->cond);
> - pthread_mutex_unlock(&cbdata->lock);
>
> +}

I think these can be static now.
> -
>
> +static DWORD WINAPI dispatch_thread(void *user)
> +{
> - struct cb_data *cbdata;
> -
> - CoInitializeEx(NULL, COINIT_MULTITHREADED);
> -
> - pthread_mutex_lock(&cb_list_lock);
> -
> - while(1){
> -
>    pthread_cond_wait(&cb_list_cond, &cb_list_lock);
> -
> -
>    while(!list_empty(&cb_list)){
> -
>        cbdata = LIST_ENTRY(list_head(&cb_list), struct cb_data, entry);
> -
>        list_remove(&cbdata->entry);
> -
> -
>        TrySubmitThreadpoolCallback(&perform_cb, cbdata, NULL);
> -
>    }
> - }
> -
> - pthread_mutex_unlock(&cb_list_lock);
> -
> - CoUninitialize();
> -
> - return 0;
>
> +}

As long as you're moving code here, can you please improve the
formatting a little? (i.e. space after "while", opening brace on a new
line.)
> - default:
> -
>    {
> -
>        ERR("Wrong callback forwarder called\n");
> -
>        assert(0);

I think this belongs in a separate patch. Also, the ERR() seems
unnecessary; the assert() gives you just as much information.