[MPlayer-users] mplayer kills with a broken frame error

[Mac Smith]

On 22-Apr-2012, at 6:08 PM, Reimar Döffinger wrote:

> On Sun, Apr 22, 2012 at 02:16:44PM +0200, Reimar Döffinger wrote:
>> On Sun, Apr 22, 2012 at 05:36:11PM +0530, Mac Smith wrote:
>>>>> also tried
>>>>> 
>>>>> MPlayer SVN-r34870-4.4.3 (C) 2000-2012 MPlayer Team
>>>>> 
>>>>> but its the same
>>>> 
>>>> Am am really certain it is _not_ the same.
>>>> That version will try reconnection once (which will fix it if the
>>>> connection broke due to pausing for a long time), though if that fails
>>>> it will give up. Retrying and thus hanging possibly forever is rarely
>>>> a good idea.
>>>> If it still fails that should mean your sever was down long enough to
>>>> cause a network timeout and then refuses a reconnect still after that.
>>>> Well, it could be the code does not handle a orderly server shutdown
>>>> well since the network timeout does not come into play then.
>>>> But you should be able to see from MPlayer's output when it tries to
>>>> reconnect.
>>> 
>>> 
>>> server was not down for a long, to simulate this i just did a "service httpd reload" and i dont see a any message of retying to connect.
>> 
>> You might want to try r34871.
>> However "service httpd reload" does _not at all_ work for simulating a
>> connection failing.
>> To my knowledge some Apache versions will incorrectly tell their clients
>> that the end of the file was reached when you shut down the server.
>> In that case the behaviour that MPlayer shows is exactly correct.
> 
> Actually, for HTTP/1.0 a closed connection means the file was fully
> transferred, so in that case MPlayer is correct in concluding the file
> is done and to exit.
> With 1.1 there _might_ be a content length field which allows to detect
> the situation.
> You could try below patch for that.
> However from my side the conclusion is with HTTP you should either use
> the graceful shutdown feature of the web server or just pull the plug,
> doing "service httpd reload" _will_ give incorrect "instructions" to the
> clients.
> --- stream/stream.c
> +++ stream/stream.c
> @@ -310,7 +310,8 @@ int stream_read_internal(stream_t *s, void *buf, int len)
> #ifdef CONFIG_NETWORKING
>     if( s->streaming_ctrl!=NULL && s->streaming_ctrl->streaming_read ) {
>       len=s->streaming_ctrl->streaming_read(s->fd, buf, len, s->streaming_ctrl)
> -      if (s->streaming_ctrl->status == streaming_stopped_e)
> +      if (s->streaming_ctrl->status == streaming_stopped_e &&
> +          (!s->end_pos || s->pos == s->end_pos))
>         s->eof = 1;
>     } else
> #endif

Thanks. I'll try this patch.