Re: Pause and resume AVI capture (an old chestnut )

Hi

Sorry to be beating a dead? horse, but , I still cant get it to
work...

Heres my fillbuffer logic...

1) Wait until its time to render the new frame ( using Advisetime for
laststreamtime + frametime (using the system clock) , I have found
this way gives the most accurate "sampling" unlike Sleep() )
2) Capture the screen
3) Timestamp with currenttime from a internal clock ( I have a private
CBaseReferenceClock member for this)
4) if the state is paused then set the delta of the private clock back
by 1 frame time ( so in effect the timestamp is constant when
paused ), keep a count of the number of "paused" frames
5) If we went from paused to running, set the clock forward by the
number of paused frames * frametime

so timestamps go monotonically upward but stay constant when paused,
then "jump" back to the real time.
or in other words, the current timestamp is the streamtime minus the
total time the graph was paused.

But what i notice is that the fillbuffer gets called regularly when
playing and paused, the moment it is resumed, i get a few calls then
it stalls ( also the internal CbaseReferenceclock doesnt like to jumb
by more than 1 second and issues a warning in the debug output ) and
thereafter the function seems to be called in bursts ... and
irregularly or completely stall.

What would a capture filter do?
( Im going to check what it would do ..)

Vivek



.

Relevant Pages

  • Re: Article: A Century of Einstein
    ... then frames won't enter into the laws of physics. ... and wrt all frames whatsoever. ... begin with systems not in the lab frame. ... As B intercepts the leading end of the platform a toggle on his clock is ...
    (sci.physics)
  • Re: Article: A Century of Einstein
    ... then frames won't enter into the laws of physics. ... and wrt all frames whatsoever. ... begin with systems not in the lab frame. ... As B intercepts the leading end of the platform a toggle on his clock is ...
    (sci.physics.relativity)
  • Re: Some Contradictory Claims in SR:
    ... ruler. ... > moving rod in the observer's frame is contracted. ... SR posits that neither the metal bar nor a clock ... NO ASSUMPTION of absolute rest is required, ...
    (sci.physics)
  • Re: Triplet Paradox
    ... according to Lorentz transformation Al's clock runs slower as ... It takes Bill ... He can do this in his outgoing frame or his ... It implies that Bill's clock run slower than Al's clock. ...
    (sci.physics.relativity)
  • Re: Clock synchronization - misinterpretation?
    ... By that I mean that when clock1 shows ... presume you agree that in this single inertial frame clock2 is ... reference frame. ... each of these four points I have one clock1 type clock, ...
    (sci.physics.relativity)
Quantcast