[FFmpeg-devel] [PATCH] add timeout to udp_read

[Hagen Schmidt]

On 07.10.2009 at 14:05, Michael Niedermayer wrote:
> On Wed, Sep 23, 2009 at 10:30:26PM +0200, Hagen Schmidt wrote:
> > On 15.09.2009 at 03:03, Michael Niedermayer wrote:
[...]
> >  static int decode_interrupt_cb(void)
> >  {
> > -    return (global_video_state &&
> > global_video_state->abort_request); +    if (timeout_network &&
> > !timeout_expired && reference_time) { +        timeout_expired =
> > av_gettime() - reference_time > timeout_network * 1000; +       
> > if (timeout_expired)
> > +            fprintf(stderr, "Timeout for receiving network
> > stream expired after %i msec\n", timeout_network); +    }
> > +    return (global_video_state &&
> > global_video_state->abort_request) || timeout_expired; }
> 
> is it neccesary to design this so complex?
> dosnt
> if (reference_time && av_gettime() - reference_time >
>  timeout_network){ fprintf(stderr, "Timeout for receiving network
>  stream expired after %i msec\n", timeout_network); return 1;
> }
> return global_video_state && global_video_state->abort_request;
> 
> work as well?
Argh, using timeout_expired is obvious senseless of course...

The time calculation
  av_gettime() - reference_time
is executed at least once by decode_interrupt_cb(). Therefore I 
disabled the calculation by
  if (timeout_network && ...
because I expect that timeout_network will be rarely set. But I prefer 
your proposal.

> here timeout_network could be set to UINT64_MAX for the disabled
>  case
Ok, done. I add opt_timeout_network() to check boundaries from 0 to  
UINT64_MAX. A problem arises while checking latter limit by 
parse_number_or_die() because the function compares the boundaries 
after a double cast. This results in an upper limit which may be 
greater than UINT64_MAX caused by the limited precision of double.
Therefore I reduced the maximum boundary to UINT64_MAX-20000 to 
guarantee always being less than UINT64_MAX after the double cast.

At the moment I have no idea to calculate the largest double value 
which is <=UINT64_MAX. I tested that the difference between the 2 
subsequent double values at UINT64_MAX is 2048 (double1 < UINT64_MAX < 
double2). Therefore UINT64_MAX-2049 could be sufficient instead of 
UINT64_MAX-20000.

The solution doesn't reject input values for the new command line 
option nto being above the upper limit between the range 
18446744073709531137 and the subsequent double value because the range 
equals 1 double value. Therefore no matter.

Many thanks for support and I'm sorry for my late response, I was 
quite busy.
Hagen
-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch.diff
Type: text/x-patch
Size: 2867 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20091115/197a8576/attachment.bin>