Do I need to do something else, for the font changes?
I checked everything out at ~530pm EST today, and the two crashes from
yesterday (one in mythtv and one in mythfilldatabase) are fixed. Thanks
Isaac!
Quick question: the OSD no longer shows how far behind live tv, when the
user pauses or rewinds... Is this a bug? The channel number and program
description show up OK. Could this have something to do with the recent
font changes?
Regarding the recent change to 'nexttrigger' calculation:
I'm going to back out this change, it was introducing a constant offset
into the 'nexttrigger'. This is why:
Suppose we trigger at time C. Then we display the frame, which takes
time X. At this point, with the change that I'm backing out, we set
nexttrigger to C+X+F where F is the frame period. Lastly we apply
feedback to correct for a fraction of the measured error. Suppose for
the sake of simplicity that no error is measured, and no feedback is
applied. Then C+X+F is the final value of nexttrigger -- but it should
be C+F if no error is measured.
This caused the video to fall behind the audio, until it was trailing by
100ms. At 100ms, additional feedback kicks in, and this was preventing
it from getting further out of sync.
I have checked in these changes, which should fix the problem:
* Nexttrigger is C+F before feedback correction
* When we are >100 ms out of sync, the feedback will correct 1/3 of
the error on the next frame (increased from 1/10). This should give
everyone very snappy resyncs.
* Increase the delay cap from 100ms to 200ms. I was seeing delays in
the 120ms range following a channel change, and we don't want to cap
those, it just slows down the resync process.
thanks
-- john
I checked everything out at ~530pm EST today, and the two crashes from
yesterday (one in mythtv and one in mythfilldatabase) are fixed. Thanks
Isaac!
Quick question: the OSD no longer shows how far behind live tv, when the
user pauses or rewinds... Is this a bug? The channel number and program
description show up OK. Could this have something to do with the recent
font changes?
Regarding the recent change to 'nexttrigger' calculation:
I'm going to back out this change, it was introducing a constant offset
into the 'nexttrigger'. This is why:
Suppose we trigger at time C. Then we display the frame, which takes
time X. At this point, with the change that I'm backing out, we set
nexttrigger to C+X+F where F is the frame period. Lastly we apply
feedback to correct for a fraction of the measured error. Suppose for
the sake of simplicity that no error is measured, and no feedback is
applied. Then C+X+F is the final value of nexttrigger -- but it should
be C+F if no error is measured.
This caused the video to fall behind the audio, until it was trailing by
100ms. At 100ms, additional feedback kicks in, and this was preventing
it from getting further out of sync.
I have checked in these changes, which should fix the problem:
* Nexttrigger is C+F before feedback correction
* When we are >100 ms out of sync, the feedback will correct 1/3 of
the error on the next frame (increased from 1/10). This should give
everyone very snappy resyncs.
* Increase the delay cap from 100ms to 200ms. I was seeing delays in
the 120ms range following a channel change, and we don't want to cap
those, it just slows down the resync process.
thanks
-- john