[MPlayer-dev-eng] Re: [PATCH] Save memory in demux_rtp.cpp

[Nico Sabbi]

Carl Eugen Hoyos wrote:
> Nico Sabbi <nicola_sabbi <at> fastwebnet.it> writes:
> 
> [...]
> 
> 
>>> void ReadBufferQueue::savePendingBuffer(demux_packet_t* dp) {
>>>+  resize_demux_packet(dp, dp->len);
>>>   // Keep this buffer around, until MPlayer asks for it later:
>>>   if (pendingDPTail == NULL) {
>>>     pendingDPHead = pendingDPTail = dp;
>>
>>dp->len is already the actual size of the packet, unless you call it 
>>with 0. Is it the case?
> 
> 
> The packet was allocated with
> dp = new_demux_packet(MAX_RTP_FRAME_SIZE);
> but the patch resizes it to its actual size.
> 
> Carl Eugen
> 
> 

new_demux_packet() and resize_demux_packet() both store the length of 
the buffer in dp->len, so at any time dp->len is the amount of memory
allocated;
resize_demux_packet(dp, dp->len); is equivalent to a no-op, as you can 
see:

inline static void resize_demux_packet(demux_packet_t* dp, int len)
{
   if(len > 0)
   {
      dp->buffer=(unsigned char *)realloc(dp->buffer,len+8);
   }
   else
   {
      if(dp->buffer) free(dp->buffer);
      dp->buffer=NULL;
   }
   dp->len=len;
   if (dp->buffer)
      memset(dp->buffer + len, 0, 8);
   else
      dp->len = 0;
}

I still don't see the usefulness of this patch, so show me if (and 
where) the dp is tampered with, please

-- 
"Without a frontend, mplayer is useless" - someone in mplayer-users