[mythtv] FYI Tuning patch
danielk at cuymedia.net
Mon Sep 26 21:47:27 UTC 2005
On Mon, 2005-09-26 at 20:24 +0200, Adam Egger wrote:
> On 9/26/05, Geoffrey Kruse <gkruse at gmail.com> wrote:
> > On Sep 24, 2005, at 4:43 PM, Isaac Richards wrote:
> > > On Saturday 24 September 2005 05:18 pm, Daniel Kristjansson wrote:
> > >
> > >>> Its harder to get consistent timings at the moment because front-
> > >>> end cpu
> > >>> load has increased (<~35% at 6910 to > 84% now) to the extent
> > >>> that the
> > >>> code in NuppelVideoPlayer::DisplayNormalFrame
> > >>> invoking "playing slower due to falling behind..." is now being
> > >>> invoked.
> r7317 with tuning-patch7 (on debian64, dvb-t) doesn't have such a high
> cpu load anymore. Only 20-25% on AMD64-3000+.
This doesn't have anything to do with the patch. Issac had tried a more
aggressive prebuffering scheme, it worked fine on his dual AMD64-4400+
but caused some problems for the rest of us. :)
> But there are still some issues:
> Pressing R to record on LiveTV only shows a black screen here:
> 2005-09-26 19:41:00.939 RemoteFile::Read() failed in RingBuffer::safe_read().
> 2005-09-26 19:41:01.164 RemoteFile::Read() failed in RingBuffer::safe_read().
Hmm, I'll look at this.
> Every recording appears now twice in the recording list. It's possible
> to press enter on both entries to start watching.
Weird, this doesn't happen for me. I'm guessing it has something
to do with StartedRecording() and FinishedRecording(), I'll try
to figure out what could cause duplicates there...
> When a scheduled recording is going to begin while I watch LiveTV it
> just leaves LiveTV. Then it's not possible to start LiveTV anymore (it
> just shows a black screen for a short moment and leaves it again).
> Unfortunately it doesn't record the scheduled show. The only thing
> that activates the scheduled show is a BE restart.
I think this is fixed in my current code. I'll test it again before
I make another patch.
> ff/rw still are strange in LiveTV, most of the time it just jumps in
> very small steps only instead of 30 seconds, sometimes the time jumps
> in 30sec steps but livetv still stays at the same position.
This is another thing unrelated to the patch. Issac wrote a more
efficient keyframe detector, but it has some problems. He's been
working on it. Running "mythcommflag", with the latest SVN, on the
broken recordings may fix it...
> Do you need any traces?
I may need something for the double recording thing, but let
me look at the code first. It may be is something obvious.
More information about the mythtv-dev