PDA

View Full Version : Socket disconnects



mvallevand
2007-02-24, 12:38 PM
I am still getting the typical random unexplained disconnects,
like this

23/02/2007 21:04:57.093 ERROR [27] MVP Timer Refresh error: An existing connection was forcibly closed by the remote host

which I hope that Martin might eventually be able to fix

As I type this I am running gbpvr in debug mode and ML2 with the setting "insane amount of logging" in the hope that something useful might be detected.
I'd be glad to help out in determining the cause of this. I'm pretty sure it is not the mvp itself, because I've streamed natively for days without a dropped connection. I know this has been a problem for the Hauppauge dongle too, but there still might be a way to work around it once we find the cause.

I'm assuming the problem occurs while media is streaming so I will run overnight with my own mvpmc debug logging. I noticed that ML2 has an option "At the End of a Random Playlist - start a new random playlist". How do I enable this?

Also If you could zip up the logs for a few minutes leading into the problem and send them that would help me too.

Martin

tkgafs
2007-02-24, 12:50 PM
I'm assuming the problem occurs while media is streaming so I will run overnight with my own mvpmc debug logging. I noticed that ML2 has an option "At the End of a Random Playlist - start a new random playlist". How do I enable this?


in the gbpvr config app go to the plugins screen and select the musiclibrary2 plugin, then choose the settings button

when in the musiclibrary2 settings screen choose the playlists/playing tab and then click on the radio button for "At the end of a random playlist ..." newplaylist

then click ok to save the settings


typically my mvp has now run for about 4 hours without fail :)

but when I do eventually get the error I'll send the relevant parts of the logs

Tkgafs

mvallevand
2007-02-24, 07:56 PM
Tkgafs, its even easier than I thought to cause this. I just boot don't touch a thing other then display the screen saver and I get an MVP Timer error after a period of time. I've just sync'd the time between the two so I can find out which process dies first, mvpmc gives the same socket error message "the peer has disconnected".

Martin

tkgafs
2007-02-24, 09:49 PM
ok after about 9 hours of ML2 playing

I got a disconnect

details as follows

U/V buffer: 1265 bytes
24/02/2007 21:13:02.109 ERROR [27] MVP Timer Refresh error: An existing connection was forcibly closed by the remote host
24/02/2007 21:13:02.468 ERROR [27] at System.Net.Sockets.Socket.Send(Byte[] buffer, Int32 offset, Int32 size, SocketFlags socketFlags)
at System.Net.Sockets.Socket.Send(Byte[] buffer)
at GBPVR.xbb1566157f5f8b1f.xcca58d4608994dd8(Point x557cda7ca079f710, Rectangle x19534874af95b268, Bitmap x84901bd355736ef6, Boolean x484012b92f24af4e)
at GBPVR.xbb1566157f5f8b1f.xe75b6be209ce111f(Image x84901bd355736ef6)
at GBPVR.xbb1566157f5f8b1f.xb1cbc922a07ef39b(Object xdf2e3583f942db7b, EventArgs xc4f45905cb1fc7ba)
24/02/2007 21:13:22.062 VERBOSE [25] MVP data taking too long, connection must have been dropped!
24/02/2007 21:13:22.062 VERBOSE [25] MVP Control Thread shutting down....

attached are extracts from the logs gbpvr.exe.log musiclibrary2.log and an event from the system event-log

tkgafs

mvallevand
2007-02-25, 02:07 PM
As an update, on this (when running just the screensaver) I think this might have something to do with the socket keepalive value. I had two crashes at 2 hours which is what the default kernel value is. I lowered the keepalive to 5 minutes and I can now crash gbpvr consistently much faster so there might be a relation.

My guess is when the mvpmc socket sends the keepalive value, gbpvr ignores it or there is a socket error, and because mvpmc doesn't get a keepalive return the socket is closed on its own.

Martin

sub
2007-02-25, 03:28 PM
Are you talking about some application defined keep alive message, or the underlying sockets 'keepalive'? Assuming the latter, this all should take card of automatically by windows behind the scenes of the .NET socket libraries. I'm not aware I'm supposed to do anything manually, correct?

mvallevand
2007-02-25, 03:52 PM
Are you talking about some application defined keep alive message, or the underlying sockets 'keepalive'? Assuming the latter, this all should take card of automatically by windows behind the scenes of the .NET socket libraries. I'm not aware I'm supposed to do anything manually, correct?

I'm talking socket level, to play around with the kernel default 7200 seconds (that is rfc value) I changed the SO_KEEPALIVE to 300 for the socket to get the fast "crash". What I did notice is that according to ethereal the last two commands that mvpmc issued had the RST reset flag on the gui stream and the FIN flag on media stream. Since the gbpvr ip ack'd the RST which is a socket close.

I'm not sure what can be done. Certainly when one searches keepalive with "An existing connection was forcibly closed by the remote host" there's almost too much to weed through since it is part of that return code. What you will see is many applications see the same behaviour, this error occurring after a period of time.

I'm going to try setting the keepalive to MAXINT and let it run overnight


Martin

mvallevand
2007-03-01, 04:01 AM
As a follow-up on this thread, I have left emulation mode running with the screen saver from the Hauppauge software for 24 hours, and there was no socket level disconnect, and it played audio and video just fine afterwards. I have emailed Hauppauge to see if they code anything special. I'm not sure where I can go, since it is not emulation mode code sending the disconnect.

Martin

tkgafs
2007-03-04, 12:20 PM
Martin,

have spent the last few days trying out the 2nd of march and earlier dongles

with any of these the best I have managed to run MusicLibrary2 for is about 2 hours 25 minutes.

I am using version 99.12 of gbpvr I have now reverted to dongle 3_3.24265b.bin downloaded from the shspvr site, at my first try ML2 has been playing now for 15 hours without a break, 19 hours is my record for this dongle.

not sure how this helps you other than to say your alpha dongle are still very unstable whilst playing music at least.

There doesnt seem to be any pattern to the failures, the gbpvr server is the family computer used for internet browsing email etc for 5 different people, its running XP professional SP2 and I have set it up fast user switching.

GBpvr is set up to not interact with the desktop.

the mvp has crashed when nobody has touched the PC for ages, it often crashes when someone else logs on.

the MVP is connected via a direct crossover cable and a 2nd ethernet card is set up for internet sharing so providing a dchp server for the MVP the ethernet card the mvp is connected to is fixed at the ip address 192.168.0.1

cant think of much else background which could possibly help

Tkgafs

mvallevand
2007-03-04, 02:12 PM
with any of these the best I have managed to run MusicLibrary2 for is about 2 hours 25 minutes.

I am using version 99.12 of gbpvr I have now reverted to dongle 3_3.24265b.bin downloaded from the shspvr site, at my first try ML2 has been playing now for 15 hours without a break, 19 hours is my record for this dongle.

not sure how this helps you other than to say your alpha dongle are still very unstable whilst playing music at least.

There doesnt seem to be any pattern to the failures, the gbpvr server is the family computer used for internet browsing email etc for 5 different people, its running XP professional SP2 and I have set it up fast user switching.

the mvp has crashed when nobody has touched the PC for ages, it often crashes when someone else logs on.


Our findings are similar, although I went 11h 40 once and it died when I woke the pc's screen saver. I haven't stopped looking.



the MVP is connected via a direct crossover cable and a 2nd ethernet card is set up for internet sharing so providing a dchp server for the MVP the ethernet card the mvp is connected to is fixed at the ip address 192.168.0.1

cant think of much else background which could possibly help

Tkgafs

If you have a chance, you're the first person I know that has actually connected via crossover, could you tell me if mvpmc's native internet radio streaming works? Use the Filesystem option and go to /usr/playlist and select default.m3u. I'm not sure how the dns or routing will work so I'm not optimistic.

Thanks,

Martin

tkgafs
2007-03-04, 06:24 PM
If you have a chance, you're the first person I know that has actually connected via crossover, could you tell me if mvpmc's native internet radio streaming works? Use the Filesystem option and go to /usr/playlist and select default.m3u. I'm not sure how the dns or routing will work so I'm not optimistic.

Thanks,

Martin

Radio streaming seems to work fine although I only tried a couple of stations which didn't say they needed realplayer or vlc

I seem to remember that resolv.conf points correctly at the other end of the crossover cable [192.168.0.1] but I didnt actually check !

Tkgafs

mvallevand
2007-03-04, 08:04 PM
Radio streaming seems to work fine although I only tried a couple of stations which didn't say they needed realplayer or vlc

I seem to remember that resolv.conf points correctly at the other end of the crossover cable [192.168.0.1] but I didnt actually check !

Tkgafs
Great thanks, one working was good enough for me. I can use this as a test that ICS/DNS, and DHCP are set up properly.

Martin

mvallevand
2007-03-11, 12:39 AM
I'm probably going to stop looking into this for now. Hauppauge tells me they do nothing on the mvp and we use the same kernel, so I think this could be a Windows issue more than mvpmc

I did find that I could greatly extend the time that mvpmc stayed connected, I got the logs rolling mesage for the first time, by adding the registry key eepAliveTime and setting it to 1800000 (half an hour).

This link has info on this setting as well as many others.

http://support.microsoft.com/kb/314053

Martin

tkgafs
2007-03-11, 10:41 AM
I'm probably going to stop looking into this for now. Hauppauge tells me they do nothing on the mvp and we use the same kernel, so I think this could be a Windows issue more than mvpmc

I did find that I could greatly extend the time that mvpmc stayed connected, I got the logs rolling mesage for the first time, by adding the registry key eepAliveTime and setting it to 1800000 (half an hour).

This link has info on this setting as well as many others.

http://support.microsoft.com/kb/314053

Martin

Martin

can I check what exactly you did as the following seems to say that the default for KeepAliveTime is 7,200,00 [two hours] so in effect you seem to be reducing the time between the keepalive transmissions

the following is extracted from the microsoft page Martin mentions above

KeepAliveInterval
Key: Tcpip\Parameters
Value Type: REG_DWORD - Time in milliseconds
Valid Range: 1 - 0xFFFFFFFF
Default: 1000 (one second)
Description: This parameter determines the interval that separates keepalive retransmissions until a response is received. After a response is received, KeepAliveTime again controls the delay until the next keepalive transmission. The connection is aborted after the number of retransmissions that are specified by TcpMaxDataRetransmissions are unanswered.

KeepAliveTime
Key: Tcpip\Parameters
Value Type: REG_DWORD - Time in milliseconds
Valid Range: 1 - 0xFFFFFFFF
Default: 7,200,000 (two hours)
Description: The parameter controls how frequently TCP tries to verify that an idle connection is still intact by sending a keepalive packet. If the remote computer is still reachable and functioning, the remote computer acknowledges the keepalive transmission. By default, keepalive packets are not sent. A program can turn on this feature on a connection.

Tkgafs

mvallevand
2007-03-11, 12:23 PM
can I check what exactly you did as the following seems to say that the default for KeepAliveTime is 7,200,00 [two hours] so in effect you seem to be reducing the time between the keepalive transmissions


I'm using the following setting right now, As for the default the documentation isn't clear since it also says "By default, keepalive packets are not sent."

========================================
Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Servic es\Tcpip\Parameters]
"KeepAliveTime "=dword:001b7740
========================================

Martin

mvallevand
2007-03-20, 07:05 PM
I suggest applying this new patch from sub. I just ran 14 hours straight with it.

http://forums.gbpvr.com/showpost.php?p=185143&postcount=31

This patch also clears up a minor annoyance on startup where mvpmc shows the screen but gbpvr thinks that it is in screen saver mode although the jump to screen saver can be abrupt.

Martin

sub
2007-03-20, 07:09 PM
I suggest applying this new patch from sub. I just ran 14 hours straight with it.

http://forums.gbpvr.com/showpost.php?p=185143&postcount=31
Thanks for the confirmation. Sometimes I wonder if anyone even downloads my patches :D

mvallevand
2007-03-20, 09:29 PM
Well I just did good thing I've backup the two files because when I run this patch I loose a great share of my RAM..
Services Running look huge
I let LiveTV run for about an hour and didn't see anything close to the those numbers. 18MB for dormant session 22-34 for the running one. Also I don't think you use mvpmc do you?

Martin

sub
2007-03-20, 09:31 PM
Services Running look huge
http://img1.putfile.com/thumb/3/7815480477.jpg (http://www.putfile.com/pic.php?img=5030725)Actually those can be fairly normal numbers if you're running any plugins. Some plugins like Weather etc consume heaps of memory. Try temporarily removing all the files from you plugins directory. This patch wont have affected this.

Lindsay
2007-03-20, 09:56 PM
Thanks for the confirmation. Sometimes I wonder if anyone even downloads my patches :D

What a sad statement:( I have downloaded the patch, along with pretty much all the others, so I hope you feel a little better:)

sub
2007-03-20, 09:58 PM
lol :D ta

tkgafs
2007-03-21, 12:41 AM
What a sad statement:( I have downloaded the patch, along with pretty much all the others, so I hope you feel a little better:)

Snap me too I download all the patches that appear vaguely relevant

I hadn't noticed this one though so will try it out tonight

tkgafs

HtV
2007-03-21, 08:34 AM
Snap me too I download all the patches that appear vaguely relevanttkgafs

So do I.:) Just to make you fell better sub;) I hope autumn isn't getting to you.:( Overhere spring is coming.:D

tkgafs
2007-03-21, 07:03 PM
Patch seems to be make a big difference, my first shot ran ML2 for 7 hours before receiving the socket forcibly closed message.

Tkgafs

mvallevand
2007-03-22, 05:31 AM
in the gbpvr config app go to the plugins screen and select the musiclibrary2 plugin, then choose the settings button

when in the musiclibrary2 settings screen choose the playlists/playing tab and then click on the radio button for "At the end of a random playlist ..." newplaylist

then click ok to save the settings

I've been trying to make this work to see if I can better your 8 hours. Does it only build a newplaylist from songs you've haven't played? My online test database is quite small as my NAS collection is now FLAC format.

Martin

tkgafs
2007-03-22, 11:37 AM
I've been trying to make this work to see if I can better your 8 hours. Does it only build a newplaylist from songs you've haven't played? My online test database is quite small as my NAS collection is now FLAC format.

Martin

No it will randomly choose from the whole library when it needs to add tracks to the random playlist, which happens when the last track in the current playlist starts to play

there is a config option to turn on this behaviour although I cant remember what it actually is as I'm at work just now.

It seems to fail when ML2 does its scan to update the library but that happens both with the hauppauge dongle and your dongle

Tkgafs

mvallevand
2007-03-23, 03:46 PM
No it will randomly choose from the whole library when it needs to add tracks to the random playlist, which happens when the last track in the current playlist starts to play

there is a config option to turn on this behaviour although I cant remember what it actually is as I'm at work just now.

It seems to fail when ML2 does its scan to update the library but that happens both with the hauppauge dongle and your dongle


I upgraded my version of ML2 and I can loop now and last night I think I was able to determine one problem when I died after four hours. When there is a transition between songs there is a period of time where I don't request any screen or media updates as I wait for the audio buffer to empty. I'm going to reduce this wait, which will mean that the playing will become closer to gap-less (I hope this isn't a problem) and the OSD won't match 100% with what is playing during transitions. With really low bit rate files, more than 30 seconds can be stored in the buffer, but with 192k this should not be too bad a problem.

Martin

tkgafs
2007-03-23, 09:06 PM
I'm going to reduce this wait, which will mean that the playing will become closer to gap-less (I hope this isn't a problem)

Martin

Martin,

I think that getting near gapless will be a great improvement rather than a problem.

I have found that the last couple of versions from 20 & 22nd march have not performed well neither managed more than about 40 mins at a time

I also found that I could crash it by pressing the volume button up and down a couple of times

Tkgafs

mvallevand
2007-03-24, 11:36 AM
Martin,

I think that getting near gapless will be a great improvement rather than a problem.

I have found that the last couple of versions from 20 & 22nd march have not performed well neither managed more than about 40 mins at a time

I also found that I could crash it by pressing the volume button up and down a couple of times


I made some pretty significant code changes this week so I am not surprised that it is not performing as well. I'm reluctant too revert some of this code back until sub can help out with the the mini guide issue. I have better debug info right now.

I've also just lowered the priority of the mvp process via the config.xml and have noticed a few differences that I want to experiment with.

Gapless does seem to be working the way I want but I want to do a little more testing before uploading it.

Martin

mvallevand
2007-03-26, 04:37 PM
I've analyzed some data from last night's run and I did encounter a situation that I can't recover from easily, because of the Hauppauge protocol. The screen updates come in compressed form and last night the zlib uncompress() of the YVU alpha channel update returned a Z_DATA_ERROR which caused mvpmc to disconnect.

I'm guessing while mp2 is playing a problem could occur if the screen update information and the RDC play command for the next song is being received on the socket together with an OSD update. I can't control this, but it now gives me something to look for in my dumps and traces.

Update: In both disconnects (7 hrs and 2 hrs) today. There was unsolicited play information from gbpvr mixed in with the OSD screen saver update. I can now continue without crashing, but I cannot respond to the play command.

Martin

mvallevand
2007-03-28, 02:46 AM
I have found that the last couple of versions from 20 & 22nd march have not performed well neither managed more than about 40 mins at a time

I have uploaded a new dongle which I think gets rid of the conflict between the unsolicited request to play the next song and the unsolicited screen saver update. I based the logic on the GBPVR screen saver update arriving at approx. 10 second intervals, although I couldn't find a config.xml setting for this

The way the logic now works, is at the end of a song, emulation mode will not sent the end of song notice if it is expecting a screen update (ie between 9 and 11 seconds) until after the screen update is received.

On my latest test I've now gone 11 hours with MP2 streaming which is the longest I've ever had. In those 11 hours the conflict happened 20 times so there was a good chance the old dongle would have closed the socket.

I ended up taking out all the gapless code but assuming this new logic works, I should be able to reinstate something.

Martin

mvallevand
2007-03-28, 09:30 PM
I've successfully run last night's dongle for thirty hours straight with ML2 , so the prime issue is definitely resolved. With 3.1 GB transfered almost 900,000 packets and I guess this is as close to a kiosk type player as I can expect.

Martin.

tkgafs
2007-03-28, 09:42 PM
I've successfully run last night's dongle for thirty hours straight with ML2 , so the prime issue is definitely resolved. With 3.1 GB transfered almost 900,000 packets and I guess this is as close to a kiosk type player as I can expect.

Martin.

Sounds good

I'm afraid I only managed a paltry 8.5 hours :) the first time but strangely only 20 mins after that restart

I haven't been able to test any more yet but will try a run overnight tonight

tkgafs

mvallevand
2007-03-28, 09:48 PM
Sounds good

I'm afraid I only managed a paltry 8.5 hours :) the first time but strangely only 20 mins after that restart

I haven't been able to test any more yet but will try a run overnight tonight

tkgafs

Did it stop while playing with ML2 or while using the remote? I've been focusing on the playing time, I expect there still could be unavoidable collsions when the remote is clicked between songs. Also when it stopped did it crash or did ML2 just stop playing music?

Martin

jim08127
2007-03-28, 09:49 PM
Not sure if this thought is helpful. Is there anyway to test the frequency of the conflicting screen saver data and adjust dynamically in your code?

tkgafs
2007-03-29, 12:03 AM
Did it stop while playing with ML2 or while using the remote? I've been focusing on the playing time, I expect there still could be unavoidable collsions when the remote is clicked between songs. Also when it stopped did it crash or did ML2 just stop playing music?

Martin

I dont actually know as I was at work, I just started it in the morning and turned the tv off, when I returned the mvp was back at the mvpmc menu I would imagine it was just playing as its unlikely my wife touched it.

when I restarted it by clicking on emulation mode it ran for about 20 mins and it was simply playing music no one touched the remote

Tkgafs

mvallevand
2007-03-29, 12:38 AM
Not sure if this thought is helpful. Is there anyway to test the frequency of the conflicting screen saver data and adjust dynamically in your code?
Because the vnc-like data is hard to distinguish the only way I could think if is to take an average for update interval ignoring those not near a keystroke. As long as it is 10 seconds this won't make too difference, it is not exactly 10 sec. anyway.

Martin

mvallevand
2007-03-29, 11:45 AM
I dont actually know as I was at work, I just started it in the morning and turned the tv off, when I returned the mvp was back at the mvpmc menu I would imagine it was just playing as its unlikely my wife touched it.

when I restarted it by clicking on emulation mode it ran for about 20 mins and it was simply playing music no one touched the remote
That's disappointing so I tried again overnight with the actual alpha and it is still running. I'll check after work

Here is the ifconfig output at the start

~ Midnight
RX packets:19135 errors:0 dropped:0 overruns:0 frame:0
TX packets:12044 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:25460368 (24.2 MiB) TX bytes:0 (0.0 B)

7 1/2 hours later

RX packets:579152 errors:0 dropped:0 overruns:0 frame:0
TX packets:379185 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:762145306 (726.8 MiB) TX bytes:0 (0.0 B)

It did crash after 14 hours

RX packets:1103519 errors:0 dropped:0 overruns:0 frame:0
TX packets:722786 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:1451916782 (1.3 GiB) TX bytes:0 (0.0 B)

But the error did happen where I think there is a problem in gbpvr (the 20 comes from me)

02:31:14.062 VERBOSE [8] Responding to ping
02:31:14.062 VERBOSE [8] getValue cached value: /settings/DisableHauppaugeMVPScreenSaver : true
02:31:14.062 VERBOSE [8] MVP request processed
02:31:14.082 VERBOSE [10] Y/Y2 buffer: 2711 bytes
02:31:14.092 VERBOSE [8] MVP data read: 1 (4)
02:31:14.092 VERBOSE [8] User pressed: 20
02:31:14.092 VERBOSE [8] MVP request processed
02:31:14.122 VERBOSE [10] U/V buffer: 1088 bytes
02:31:14.122 VERBOSE [10] incremental frame sent
02:31:14.122 VERBOSE [10] MVP async queue items: 1
02:31:14.122 VERBOSE [10] MVP async queue handling key: 20
02:31:24.067 ERROR [10] MVP Timer Refresh error: An established connection was aborted by the software in your host machine
02:31:24.087 ERROR [10] at System.Net.Sockets.Socket.Send(Byte[] buffer, Int32 offset, Int32 size, SocketFlags socketFlags)
at System.Net.Sockets.Socket.Send(Byte[] buffer)
at GBPVR.xbb1566157f5f8b1f.xcca58d4608994dd8(Point x557cda7ca079f710, Rectangle x19534874af95b268, Bitmap x84901bd355736ef6, Boolean x484012b92f24af4e)
at GBPVR.xbb1566157f5f8b1f.xe75b6be209ce111f(Image x84901bd355736ef6)
at GBPVR.xbb1566157f5f8b1f.xb1cbc922a07ef39b(Object xdf2e3583f942db7b, EventArgs xc4f45905cb1fc7ba)
2007-03-29 02:32:24.094 VERBOSE [8] MVP data taking too long, connection must have been dropped!


Martin

tkgafs
2007-03-30, 07:29 AM
Hi Martin,

I'm still getting the disconnects the following one is after about 5 hours

2007-03-29 11:18:52.093 WARNING [23] Received unknown request of type: 133
2007-03-29 11:18:52.093 WARNING [23] Received unknown request of type: 224
2007-03-29 11:18:52.093 WARNING [23] Received unknown request of type: 0
2007-03-29 11:18:52.093 WARNING [23] Received unknown request of type: 0
2007-03-29 11:19:43.203 ERROR [25] MVP Timer Refresh error: An established connection was aborted by the software in your host machine
2007-03-29 11:19:44.078 ERROR [25] at System.Net.Sockets.Socket.Send(Byte[] buffer, Int32 offset, Int32 size, SocketFlags socketFlags)
at System.Net.Sockets.Socket.Send(Byte[] buffer)
at GBPVR.xbb1566157f5f8b1f.xcca58d4608994dd8(Point x557cda7ca079f710, Rectangle x19534874af95b268, Bitmap x84901bd355736ef6, Boolean x484012b92f24af4e)
at GBPVR.xbb1566157f5f8b1f.xe75b6be209ce111f(Image x84901bd355736ef6)
at GBPVR.xbb1566157f5f8b1f.xb1cbc922a07ef39b(Object xdf2e3583f942db7b, EventArgs xc4f45905cb1fc7ba)
2007-03-29 11:20:46.359 INFO [2] UbuRadio destructor called


do you get lots of messages similar to

WARNING [23] Received unknown request of type: XXX

where XXX is a number

in the gbpvr logs, I have no idea what the messages are or what generates them but they are always there

I have attached the relevant logfile although I dont think it is of much interest

Tkgafs

mvallevand
2007-03-30, 01:43 PM
Hi Martin,

I'm still getting the disconnects the following one is after about 5 hours

I will keep plugging away at this and getting dumps. It is tougher testing for me since it typically will take more than 8 hours. My older gbpvr machine is having trouble saving BIOS setting so this is saving me re-entering it anyway.



do you get lots of messages similar to

WARNING [23] Received unknown request of type: XXX

where XXX is a number

in the gbpvr logs, I have no idea what the messages are or what generates them but they are always there



Im prettty sure this error is a collision between the heart beat signal and the gui meesage indicating the progress within a song. I see these with the Hauppauge dongle too and its sort the of reverse to what I think is crashing emulation mode. I don't know how to avoid them or if this causes a problem for gbpvr.

Thanks,

Martin

tkgafs
2007-03-30, 06:29 PM
I will keep plugging away at this and getting dumps. It is tougher testing for me since it typically will take more than 8 hours. My older gbpvr machine is having trouble saving BIOS setting so this is saving me re-entering it anyway.



Martin,

if there are any dumps I can take let me know, as I'm keen for you to nail the problem

Tkgafs

mvallevand
2007-03-30, 06:45 PM
if there are any dumps I can take let me know, as I'm keen for you to nail the problem

Would you be able to run mvpmc via telnet from a share? I've got both the mvpmc debug version and an extreme debug version which I've never distributed if you can. Otherwise I could also distribute a custom dongle.

Speaking of that I did just find another both out of sync error using the extreme debug version which I hope to avoid in the next version.

Here is the sample of the extreme debug I've actually bolded the error, but I'm not expecting anyone to understand it.

14:04:17 YUV2 update 30 0 640 480
RE 1 :00
RE 3 :cc 00 01
RE 12 :00 1e 00 00 02 80 01 e0 00 00 00 09
encoding 9
RE 4 :00 00 a0 82
RE 2 :00 00
14:04:18 Timer 0 Ping - 5 5000

RE 2 :00 02
RE 4 :08 00 00 00
RE 2 :62 1b
RE 2 :00 00
mvpmc: read: Bad address
Error updating screen
Read 25117 extra
.

Martin

tkgafs
2007-03-30, 10:48 PM
Would you be able to run mvpmc via telnet from a share? I've got both the mvpmc debug version and an extreme debug version which I've never distributed if you can. Otherwise I could also distribute a custom dongle.

Speaking of that I did just find another both out of sync error using the extreme debug version which I hope to avoid in the next version.

Here is the sample of the extreme debug I've actually bolded the error, but I'm not expecting anyone to understand it.

14:04:17 YUV2 update 30 0 640 480
RE 1 :00
RE 3 :cc 00 01
RE 12 :00 1e 00 00 02 80 01 e0 00 00 00 09
encoding 9
RE 4 :00 00 a0 82
RE 2 :00 00
14:04:18 Timer 0 Ping - 5 5000

RE 2 :00 02
RE 4 :08 00 00 00
RE 2 :62 1b
RE 2 :00 00
mvpmc: read: Bad address
Error updating screen
Read 25117 extra
.

Martin

Yes I presume so

what do I have to put the binary into the share
telnet to the mvp and then run the binary

if thats all thats not a problem

just let me know

tkgafs

mvallevand
2007-03-30, 10:56 PM
Yes I presume so

what do I have to put the binary into the share
telnet to the mvp and then run the binary

if thats all thats not a problem

just let me know

tkgafs

Basically yes follow these instructions, until I update the sticky.

http://forums.gbpvr.com/showpost.php?p=186682&postcount=37

Thanks a lot. Note that I just uploaded a change to this file to test ski44's wireless connection. The dongle still needs a bit more testing.

Thanks again,

Martin

tkgafs
2007-03-31, 12:09 AM
Martin,
I cant get the binary to run attached is the putty log

the dongle I have loaded is the one from 28th March 22:17

Tkgafs

mvallevand
2007-03-31, 12:21 AM
Martin,
I cant get the binary to run attached is the putty log

the dongle I have loaded is the one from 28th March 22:17

Tkgafs
I looks like you forgot this

killall mvpmc

Here are some useful commands for playing in telnet.

Windows Linux (Busybox)
tasklist ps
ipconfig ifconfig
netstat netstat (-tn is fastest)

Martin

mvallevand
2007-03-31, 11:20 AM
Thanks a lot. Note that I just uploaded a change to this file to test ski44's wireless connection. The dongle still needs a bit more testing.

I've run the dongle overnight, basically since I uploaded mvpmc.zip and it is still running. It's using too much
cpu so it's not ready for upload.

Martin

tkgafs
2007-04-01, 12:51 AM
Ok Martin,

what is different between the debug binary and the dongle dated 28th March

first go at running your debug binary and it has now been playing 15 hours without a problem

I'll leave it overnight and see what happens

Tkgafs

mvallevand
2007-04-01, 01:12 AM
Ok Martin,

what is different between the debug binary and the dongle dated 28th March

first go at running your debug binary and it has now been playing 15 hours without a problem

I'll leave it overnight and see what happens

Tkgafs The main difference is that I added logic to not send one of the Hauppauge protocol heartbeats near when I'm expecting a screen saver update, you can see messages like avoid screensaver in the debug logs. It seems stable for audio, and I've fixed the CPU usage problem by removing a floating point call. Unfortunately emulation mode video isn't handling the OSD logic at all so I don't want to upload the corresponding dongle yet.

Martin

tkgafs
2007-04-01, 04:17 PM
Hi Martin,

it finally failed after about 27 hours of playing music from ML2

It was playing quite happily until I pressed "go" "go" on the remote to see what a song was called

it stayed on the screen saver for what seemed a very long time then flashed up the correct album art for a second then crashed

attached is the end of the putty log

Tkgafs

mvallevand
2007-04-01, 06:25 PM
Hi Martin,

it finally failed after about 27 hours of playing music from ML2

It was playing quite happily until I pressed "go" "go" on the remote to see what a song was called

it stayed on the screen saver for what seemed a very long time then flashed up the correct album art for a second then crashed

attached is the end of the putty log

Tkgafs

Thanks, that helps a bit. I figured it has to do with the ping logic, the latest alpha has slightly better logic, this clearly shouldn't have run the avoid logic.
But I have to do better when a key is pressed. Out of curiousity why Go Go and not Ok?

Terje Rypdal maybe?

Martin

tkgafs
2007-04-01, 07:15 PM
Thanks, that helps a bit. I figured it has to do with the ping logic, the latest alpha has slightly better logic, this clearly shouldn't have run the avoid logic.
But I have to do better when a key is pressed. Out of curiousity why Go Go and not Ok?

Terje Rypdal maybe?

Martin

yes Terje Rypdal a song from waves can't remember which now

I had pressed go go several times over the 27 hours

I press go go because I would have pressed go on the hauppage dongle
I seem to remember pressing ok caused a problem when the screen saver was on although not sure what now

tkgafs

tkgafs
2007-04-02, 08:12 AM
Hi Martin,

I've been trying out your latest dongle [3rd one from 1st of April]

it seems very unstable if you select tracks from the ML2s library to play it crashes with ease, but seems to be pretty stable if you just leave to select randomly

tkgafs

mvallevand
2007-04-02, 12:17 PM
Hi Martin,

I've been trying out your latest dongle [3rd one from 1st of April]

it seems very unstable if you select tracks from the ML2s library to play it crashes with ease, but seems to be pretty stable if you just leave to select randomly

tkgafs

I tried briefly and I couldn't get it to crash, as I queued and removed files. I will try adding album art to see if that has an impact. In this case I'd like to see your verbose gbpvr logs.

Ok, I think there was a GBPVR heart beat conflict possible on keystrokes too. I have updated the dongle to handle this.

Martin