Hello,
I have a problem with h263-1998 video plugin on windows platform, I compiled ffmpeg + opal plugins on mSys/Mingw.
First I got a crash, trace indicates
h263-1998.cxx:761 if (packetizer.Open(srcRTP.GetTimestamp(), encodedLen) < 0) { ...
Strangely it works when I add a Sleep(2) before this line. Is packetiser.Open needs a callback function which is not yet invoked by FFMPEGLibraryInstance.AvcodecEncodeVideo at this stage ?!! If that is true, shouldn't put a mutex just before and somewhere in the callback?!
However, the received first frame is ok then colours start to melt down until forming a kind of colour lines corresponding to video first pixels. a small part of the upper region is fine. I'm still unsure if that happened only because of Sleep! Do you have an idea?!
Thanks
From: s.horne@packetizer.com To: unazona@hotmail.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution Date: Tue, 19 Jan 2010 06:57:41 +1000
Carlos
Well caught. I have applied a fix into the CVS.
Simon
From: Carlos Haj [mailto:unazona@hotmail.com]
Sent: Tuesday, 19 January 2010 12:32 AM To: s.horne@packetizer.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution
Hello Simon,
Thanks for your time and effort. I had the same problem using H263 plugin on both windows and linux.
I found the glitch finally in h323plus code, the following function calls fine but sends absolutly nothing to the video plugin DLL
h323pluginmgr.cxx::730 PBoolean SetCodecControl(...):
you can verify that by tracing "encoder_set_options" in the video plugin dll!! Codec options can't be changed during a call, including framewidth, bandwidth, quality, etc...
on the other hand, the call to encoder_set_options in H323PluginVideoCodec constructor works, that's why the plugin has initial default profile values. h323pluginmgr.cxx::1590
PluginCodec_ControlDefn * ctl = GetCodecControl(codec, SET_CODEC_OPTIONS_CONTROL);
etc...
I modified the SetCodecControl static function, here is it
static PBoolean SetCodecControl(const PluginCodec_Definition * codec,
void * context,
const char * name,
const char * parm,
const char * value) { PluginCodec_ControlDefn * codecControls = GetCodecControl(codec, name); if (codecControls == NULL) return FALSE;
PStringArray list; list += parm; list += value; char ** _options = list.ToCharArray(); unsigned int optionsLen = sizeof(_options);
bool result = (*codecControls->control)(codec, context, SET_CODEC_OPTIONS_CONTROL, _options, &optionsLen);
free(_options);
return result;
/* old Code
....
char const * options[2] = { parm, value }; unsigned optionsLen = 2;
return (*codecControls->control)(codec, context, SET_CODEC_OPTIONS_CONTROL, options, &optionsLen); */ }
With that, I'm able to have a CIF-QCIF and QCIF-CIF calls!!
Hope this was helpful, good luck appreciate your effort.
Greetings, Carlos
From: s.horne@packetizer.com To: unazona@hotmail.com Subject: RE: [h323plus] Video Resolution Date: Mon, 18 Jan 2010 22:50:42 +1000
Carlos
The framesize information is passed to the encoder as part of the frame header information. The codec should adjust the frame size according to the header information passed to it. This would be a bug in the H.261 encoder.
Simon
From: h323plus-bounces@lists.packetizer.com [mailto:h323plus-bounces@lists.packetizer.com] On Behalf Of Carlos Haj Sent: Monday, 18 January 2010 9:19 PM To: h323plus@lists.packetizer.com Subject: Re: [h323plus] Video Resolution
Simon,
I've been working on this, I have more detailed results.
First of all, my webcam VideoInputDevice has the correct qCIF size before the attachVideoReader, while tracing the read of the codec it the rawChannel and grabwidth and grabheight are correct.
I traced the plugin DLL.
When H323plus callback OpenVideoChannel, the videoCodec framesize is already set using H261EncoderContext::SetFrameSize(352,288); which corresponds to default frame size I suppose.
The only problem.. A further call to codec.SetFrameSize sets the new framesize only in h323 capabilities, and it never reachs the plugin DLL I have traced one call to h261encodecontext::SetFrameSize.
No trace for h261 setframesize while invoking the codec.setframesize.
So, each call of H261 EncodeFrames(...) returns 0 on TRACE(1,"H261\tPayload of grabbed frame too small for full frame"); as assigned frameSize is different than the videoDevice.
It seems that mediaFormat.SetOptionInteger never tell the pluginDLL about the frameSize option changing.
Have you and idea?
Regards, Carlos
From: s.horne@packetizer.com To: unazona@hotmail.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution Date: Sat, 16 Jan 2010 04:05:25 +1000
Carlos
Maybe I should rephase my last email The old openH323 behaviour in the OpenVideoChannel allowed the user to safely set the webcam framesize to the default size of the codec because it was always fixed. So if your codec was CIF webcam was CIF. So that is why QCIF/CIF in both directions works. Now you have changed the frame size of CIF capability to QCIF in the codec it won't. Why? You must make sure in your code the framesize coming out of the VideoInputdevice when you create it is set to QCIF otherwise it will be CIF and the encoder will expect QCIF and that is why you are getting the codec encoder call returning 0 and the channel being closed down.
Did I miss something?
Simon
From: Carlos Haj [mailto:unazona@hotmail.com]
Sent: Saturday, 16 January 2010 3:24 AM To: s.horne@packetizer.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution
Simon,
I used fixed width and height to determine display and reader working, and I traced again my application.
This time I went deeper into the h323plus library, I think it's a bug of h323plus, between the library and the video codec plugin.
Bug Scenario:
I have set
AddAllCapabilities(0, P_MAX_INDEX, "*");
I only use H261 this time
SetVideoFrameSize(H323Capability::cifMPI);
Set the maximum Frame size and send a kind of parameter cifTag to the plugin
on my OnVideoChannel I initialise
if (encoder)
codec.SetFrameSize(176, 144);
This is the call stack by chronological order for an h323 video call using a CIF<->QCIF endpoint (trace is for endpoint that sending qcif)
1) H323PluginVideoCodec::H323PluginVideoCodec(...)
it creates the encoder, and set frame size to CIF (the default plugin maximum value)
2) H323PluginVideoCodec::SetFrameSize(...)
it adjusts the frame size, called from my OnVideoChannel, with qcif Value
3) H323PluginVideoCodec::Read(...)
All work, my rawChannel, renderer, etc... until I reach line 1690
H323pluginmgr.cxx::1690 - THE FIRST READ call to encode video fails
int retval = (codec->codecFunction)(codec, context,
bufferRTP.GetPointer(), &fromLen,
dst.GetPointer(), &toLen,
&flags);
retval = 0, which results into closing the channel as it returns false,
if I change SetVideoFrameSize(H323Capability::cifMPI); into SetVideoFrameSize(H323Capability::qcifMPI);
The codec function works again, but naturally with qCIF on both sides
I think that once the codec parameter was tagged to use a cif or a qcif it can't change resolution during the call. so the encoding function fail for a resolution different than one setted using the SetVideoFrameSize.
It's likely that the SetVideoFrameSize send an additional information to plugin code using a setcontrolcodec which are required later so the encoding function works properly.
I noticed, that video resolution fail is only a local behaviour, and has nothing to do with remote endpoint.
I will be debugging more, I still don't know the plugin code so much, so you might have a clearer idea than me? I think if you make a simple test you would encounter the same behaviour!
Thanks
From: s.horne@packetizer.com To: unazona@hotmail.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution Date: Sat, 16 Jan 2010 01:37:07 +1000
Carlos
You cannot know what the remote will send when you call OpenVideoChannel for the decoding side, that can only be done when you receive the first RTP packets.
You need to do a level 6 trace to work out who and why the channel is being closed. I don't think it is a capability issue, usually the problem comes from either the video encoder device (webcam) or the display device. Check to see that you have set the video input device to send at the same size as your encoder. The old openH323 used to do be able to do that automatically as the frame size for the capability was known. You have to make sure the input to the encoder from the webcam is correct.
Version of h323plus would be irrelevent in this case.
Simon
From: Carlos Haj [mailto:unazona@hotmail.com]
Sent: Saturday, 16 January 2010 1:06 AM To: s.horne@spranto.com.au; s.horne@packetizer.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution
Simon,
Thanks for valuable information. What is happening literally in a minimised case: I derived OnVideoChannel
on the Local endpoint
if (encoding) {
codec.SetFrameSize(352, 288); ..... } else { cout << codec.GetWidth(); cout << codec.GetHeight(); // I receive CIF (ERROR) ..... }
on the Remote endpoint
if (encoding) {
codec.SetFrameSize(176, 144); ..... } else { cout << codec.GetWidth(); cout << codec.GetHeight(); // I receive CIF (normal) ..... }
I looked into H323 trace Log, (remote endpoint which supposed to send QCIF)
PLUGIN Error encoding frame from plugin H.261-CIF H245 Request CloseLogicalChannel H323RTP Transmit H.261-CIF thread ended
It seems that the H.261 channel is closed prematurely. As my local endpoint received an incorrect video size I suppose that local endpoint causes remote to close the transmission channel.
Could this be related to the version of the plugin or h323plus lib which I use? it's likely that the video resolution is sent before codec.SetFrameSize is invoked!
Did you in practice used different video resolution under the same capability? e.g. CIF<->QCIF calls
Carlos
From: s.horne@spranto.com.au To: unazona@hotmail.com; s.horne@packetizer.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution Date: Fri, 15 Jan 2010 23:14:23 +1000
Carlos
The current behaviour is correct.
The old OpenH323 used to only send CIF with *-CIF{sw} capability and only QCIF with *-QCIF{sw} . This was changed with the video plugins so the capability could do up to CIF so it would also supported QCIF. This is perfectly H.245 legal.
In this way only 1 capability can do both frame sizes. This is the way most manufacturers implement the capability exchange. The capability has a flag for the size it supports so the old CIF capability used to say QCIF - 0 CIF - 1. So if you wanted to do QCIF you had to use the *-QCIF{sw} capability as the QCIF flag in the *-CIF{sw} was 0. This was changed so that the *-CIF{sw} capability had the QCIF flag so it could also support QCIF. Then it was safe to remove the *-QCIF{sw} capability and only exchange using the *-CIF{sw}.
Now SetVideoFrameSize() will set the global frame size limit so regardless of the direction that will be the maximum size permitted (if CIF only up to CIF capability is available). The codec.FrameSize() in the MyH323Endpoint::OpenVideoChannel() allows the implementor to set the actual frame size to send.
Is there something else I missed?
Simon
From: Carlos Haj [mailto:unazona@hotmail.com]
Sent: Friday, 15 January 2010 8:39 PM To: s.horne@packetizer.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Video Resolution
Simon,
Thanks for your quick answer, just noticed that I had a bad thread name;)
Well, I have the codec.SetFrameSize(w, h); in my OpenVideoChannel
When adding capabilities I use
AddAllCapabilities(0, P_MAX_INDEX, "*"); SetVideoFrameSize(H323Capability::cifMPI);
Capabilities table shows as
Set 0: 0: audio 1: H.261-CIF{sw}
I receive an OpenVideoChannel for encoder, I open channel and attach video reader + colourConverter, just after that I receive a close video signal from a H245Negotation, here I join Call Stack
testApp.exe!mmDDSVideoOutputBridge::close2() Ligne
359 C++
testApp.exe!mmVideoInputDevice::~mmVideoInputDevice() Ligne 43 + 0x1b C++
testApp.exe!mmVideoInputDevice::`scalar deleting destructor'() + 0x2b C++
testApp.exe!PVideoChannel::CloseVideoReader() Ligne 203 + 0x24 C++
testApp.exe!PVideoChannel::Close() Ligne 145 C++
testApp.exe!H323Codec::CloseRawDataChannel() Ligne 488 + 0x11 C++
testApp.exe!H323VideoCodec::Close() Ligne 747 C++
testApp.exe!H323Channel::CleanUpOnTermination() Ligne 716 C++
testApp.exe!H323_RTPChannel::CleanUpOnTermination() Ligne 1160 C++
testApp.exe!H245NegLogicalChannel::Release() Ligne 1219 C++
testApp.exe!H245NegLogicalChannel::HandleCloseAck(const H245_CloseLogicalChannelAck & __formal={...}) Ligne 1086 C++
testApp.exe!H245NegLogicalChannels::HandleCloseAck(const H245_CloseLogicalChannelAck & pdu={...}) Ligne 1372 + 0xf C++
testApp.exe!H323Connection::OnH245Response(const H323ControlPDU & pdu={...}) Ligne 3293 + 0x22 C++
testApp.exe!H323Connection::HandleControlPDU(const H323ControlPDU & pdu={...}) Ligne 3189 + 0x12 C++
testApp.exe!H323Connection::HandleControlData(PPER_Stream & strm={...}) Ligne 3170 + 0x18 C++
testApp.exe!H323Connection::HandleReceivedControlPDU(bool readStatus=true, PPER_Stream & strm={...}) Ligne 3103 + 0x12 C++
testApp.exe!H323Connection::HandleControlChannel() Ligne 3085 + 0x13 C++
testApp.exe!H245TransportThread::Main() Ligne 700 C++
If I remove the "SetVideoFrameSize(H323Capability::cifMPI);", I get the following capabilities set
1: H.261-CIF{sw}
H.261-QCIF{sw} H.261{sw}
with exactly same strange behaviour, QCIF is not sending while CIF is being received, by the way qcif endpoint says
Local Video Codec - H.261-CIF{sw} Local Resolution - 176x144
Remote Video Codec - H.261-CIF{sw} Remote Resolution - 352x288
I tried a "ReorderCapabilities(PStringArray("QCIF"));" after my addcapabilities, I ended up with both endpoints using QCIF and ignoring my SetFrameSize anyway!!
Is there a glitch using H323Plus video plugins? my code works perfectly on openh323 stack.
"SetVideoFrameSize(H323Capability::cifMPI);" is supposed to define a maximum video resolution?! as you said in API doc "... to set the maximum framesize allowed to the specified value "
Is that usual to have H.261-CIF with a QCIF resolution? On which order codec plugins can be added to avoid such problems of negotiation?
Regards
From: s.horne@packetizer.com To: unazona@hotmail.com; h323plus@lists.packetizer.com Subject: RE: [h323plus] Valid sessionID in OpenLogicalChannel command Date: Fri, 15 Jan 2010 08:12:54 +1000
Carlos
In your derived H323EndPoint::OpenVideoChannel(()
(isEncoding)
codec.SetFrameSize(w,h);
Simon
From: h323plus-bounces@lists.packetizer.com [mailto:h323plus-bounces@lists.packetizer.com] On Behalf Of Carlos Haj Sent: Friday, 15 January 2010 12:59 AM To: h323plus@lists.packetizer.com Subject: Re: [h323plus] Valid sessionID in OpenLogicalChannel command
Hello Guys,
I would like to have different frame sizes for two endpoints, first I tried to play around with H323Endpoint::SetVideoFrameSize, but it seems that part changes the maximum allowed size, which means that if I switch the videoFrameSize to qCIF, I wouldn't be able to receive CIF from the other endpoint. my test only received QCIF.
I tried also to modify the OpalMediaFormat by using a GetWritableMediaFormat for a capability, that ended with unusual CIF format while having a QCIF resolution.
On the other hand, I looked Opal examples where they use an OpalManager, but unfortuntly that seems incompatible with h323plus, while old openh323 tutorials and exampls don't have any video plugins support.
How I can manage to receive and send different resolutions (frame sizes) using H323plus?
I appreciate a lot your help Carlos
Discute avec tes amis partout, grâce à Messenger sur ton mobile. Cliquez ici !
Avec Internet Explorer, surfez en toute discrétion sur internet Cliquez ici !
Avec Internet Explorer, surfez en toute discrétion sur internet Cliquez ici !
Discute avec tes amis partout, grâce à Messenger sur ton mobile. Cliquez ici !
Avec Internet Explorer, surfez en toute discrétion sur internet Cliquez ici ! _________________________________________________________________ Téléchargez Internet Explorer 8 et surfez sans laisser de trace ! http://clk.atdmt.com/FRM/go/182932252/direct/01/