Re: Update: Progress and puzzles

2016-08-22 Thread RS

From: Jim web
Sent: Monday, August 22, 2016 09:36



This morning I could not get either the hls *or* 'flash' fetching to work
correctly (Still with gip 2.95 as yet.). I can't recall the pids but one
example was the Dad's Army from yesterday.


There is no hlshd or flashhd mode for Dads Army The Captains Car (b00789gm) 
last broadcast at 1930 yesterday (although I'm not sure where).

hlsvhigh1 had 3 Segment not found errors.
flashvhigh1 at 832x468 25fps was fine.
hvfvhigh1 at 960x540 25fps was also fine.





___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Re: Update: Progress and puzzles

2016-08-22 Thread RS

From: RS Sent: Monday, August 22, 2016 13:26


 Vangelis has pointed out that hvfhd2 is usually faster, so it may be 
better to use --tvmode=hvfhd2,hvfhd1


Sorry I meant to say Vangelis has pointed out hvfvhigh2 is usually faster 
than hvfvhigh1.




___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Re: Update: Progress and puzzles

2016-08-22 Thread RS

From: michael norman Sent: Monday, August 22, 2016 11:53


I have no idea if this helps but I've just installed gip 2.96 and the 
latest FFMPEG on Linux Mint 17.3 (from the PPA's)


Trying both the Dads Army Episode and the Gergieve prom with in the latter 
case get-iplayer --get 210 tvmode=hlssstd1 I get it in hfdhd1 which works 
fine here in VLC as as mp4 file. Trying hlshd1 fails.  FWIW I am quite 
happy with hfdhd1, I don't know why hls is or might be better, is it 
something to do with the sound bitrate ?


I think you may have an extra s in hlssstd1.  It should be hlsstd1.  For 
hfdhd1 I think you mean hvfhd1.  In general it is preferable to omit the 1 
or 2 suffix.  If you specify hvfhd it will try hvfhd1 first and if that 
fails it will try hvfhd2.  Vangelis has pointed out that hvfhd2 is usually 
faster, so it may be better to use --tvmode=hvfhd2,hvfhd1


The reason HLS is preferable to HVF is that it is faster and also has a 
1280x720 25fps mode, hlshd.  hvfhd is 1280x720 at 50fps, so the bit rate is 
more than twice as high (5.5Mbit/s instead of 2.5Mbit/s).  The best 25fps 
HVF resolution is 960x540, hvfvhigh.  hvfhd and hvfsd have a higher sound 
bit rate than hlshd.





___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Re: Update: Progress and puzzles

2016-08-22 Thread RS

On 22/08/16 09:36, Jim web wrote:

I have go do something else this morning but this prompts me to ask if
others are:

A) Still having missing segments from hls



B) now also having trouble with ye olde flash.


Fake or Fortune (b07r1yfh) from last night, downloaded today, had 26 Segment 
not found errors in hlshd1 in a one hour programme.

flashhd1 was fine.

BBC Business Live (b07ps51r) from this morning had 17 Segment not found 
errors in hlshd1 in a half hour programme.

flashhd1 was fine.

After I had downloaded Fake or Fortune the BBC website and iPlayer website 
seemed very slow.  The Business Live hlshd1 download kept stopping and 
averaged about 6Mbit/s instead of 60Mbit/s.  Both flashhd downloads seemed 
to be at normal speed (which is a lot slower than hlshd).


We have speculated that Akamai has applied a fix which has not worked, and 
has made things worse.  It could be coincidence that things have got worse. 
dinky's pre-Olympics list shows that originally only a few programmes were 
affected, and there might be 1 Segment not found error in the hlshd1 mode of 
a programme and with one exception not in any other HLS mode, and in about 
half the cases the error was corrected in about a week.  More recently 
almost every programme has been affected, the problem has spread to other 
HLS modes, the number of Segment not found errors has dramatically 
increased, and is increasing still further as days go by.


To me those are the symptoms of a hardware failure.  It is difficult to 
imagine though how it could be a hardware failure.  Both the BBC and Akamai 
must have heavily replicated hardware with changeover switches so that 
hardware is used in rotation.  Even at my end I have made tests which I am 
reasonably certain have not had any common hardware except for my Point of 
Presence.







___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Re: Update: Progress and puzzles

2016-08-22 Thread michael norman

Jim

I have no idea if this helps but I've just installed gip 2.96 and the 
latest FFMPEG on Linux Mint 17.3 (from the PPA's)


Trying both the Dads Army Episode and the Gergieve prom with in the 
latter case get-iplayer --get 210 tvmode=hlssstd1 I get it in hfdhd1 
which works fine here in VLC as as mp4 file. Trying hlshd1 fails.  FWIW 
I am quite happy with hfdhd1, I don't know why hls is or might be 
better, is it something to do with the sound bitrate ?


Mike

On 22/08/16 09:36, Jim web wrote:

This morning I could not get either the hls *or* 'flash' fetching to work
correctly (Still with gip 2.95 as yet.). I can't recall the pids but one
example was the Dad's Army from yesterday.

hvfhd worked fine for the sectional files of the Gergiev prom.

I have go do something else this morning but this prompts me to ask if
others are:

A) Still having missing segments from hls

B) now also having trouble with ye olde flash.

If I don't get an email this morning I'll try nudging the elbow of someone
at the BBC above these issues. My impression is that there are still
problems, but I've been focussing on getting Proms by means that seem OK.

Jim




___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Re: Update: Progress and puzzles

2016-08-22 Thread Jim web
This morning I could not get either the hls *or* 'flash' fetching to work
correctly (Still with gip 2.95 as yet.). I can't recall the pids but one
example was the Dad's Army from yesterday.

hvfhd worked fine for the sectional files of the Gergiev prom.

I have go do something else this morning but this prompts me to ask if
others are:

A) Still having missing segments from hls

B) now also having trouble with ye olde flash.

If I don't get an email this morning I'll try nudging the elbow of someone
at the BBC above these issues. My impression is that there are still
problems, but I've been focussing on getting Proms by means that seem OK.

Jim

-- 
Electronics  http://www.st-and.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
Armstrong Audio  http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc  http://www.audiomisc.co.uk/index.html


___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer


Update: Progress and puzzles

2016-08-20 Thread Jim Lesurf
Another try this morning confirmed that using the --ffmpeg option lets me
have gip successfully fetch via hvfhd and convert the result into playable
mp4 files. I also spent some time with ffprobe comparing previous hvfhd
fetch results as follows.

My earliest hvfhd fetches mostly were 'silent' when played with VLC. One or
two played. The distinction seems now to be as follows:

The few where VLC played the audio had ffprobe audio stream reports like

aac (LC) mp4a/0x6134706D fltp 125k

The ones that were 'silent' *lacked* the  (LC).

The versions I then got with the --raw option as ts files gave me

aac (LC) ... etc - i.e. as the above examples where VLC *did* play the
audio. But these have ther audio as stream 0,0 whereas it is 0,1 on the
converted files above.

Examining the later mp4s converted with the new ffmpeg confirmed the
pattern that when the report included the (LC) then VLC could play the
audio, but was silent when the (LC) was absent.

Quite how this squares with what I was told by my BBC contacts I don't
know. When I find out more I may be able to resolve that.

Following on from that, a curio...

A few days ago I used Flash mode to get the old 1280x720 25fps version of
the Beriloz Romeo and Juliet (Prom 20). So I thought I'd try getting the
hvfhd version as well for comparison. A quick check on the main Proms TV
list (URL given in a previous email) showed that now Prom 20 was absent
from the list! So I could not check for any 'part' files. And a newcomer
might decide the prom wasn't available.

However the pid I'd already used (b07mlqhm) duly got me the hvfhd 50fps
version and it seems fine. Although during the end conversion gip/ffmpeg
gave a series of reports specifying a sort of 'how I am getting on' which
I've not noticed before. This gave a set of elapsed times - 4m11s 8m7s,
etc. Thinking of the 'missing segments' I have a look at playing though
some of these. Only did a few, but didn't notice any problems. So maybe
these are some kind of progress reports from ffmpeg now?

BTW Can someone say what settings/settings gip has ffmpeg use when turning
the hvfhd result from ts into mp4? Using a verbose output didn't show this
when I used gip, although I may have missed it.

Once I know,  I can expriment with it on the ts files I got earlier using
--raw to see if that creates playable mp4 that let VLC give sound OK.

With luck, this all means that when the Lumpits end and we get a flood of
delayed proms txs we can get good versions of them. Thanks again to
dinkypumpkin for making this possible! At present I doubt I'll want the
1280x720 50fps for all files. But they do improve the proms over previous
'lower spec' streams! :-)  Just a shame the BBC couldn't go to 320k aac for
the audio 8-]

Jim

-- 
Electronics  http://www.st-and.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
Armstrong Audio  http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc  http://www.audiomisc.co.uk/index.html


___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer