Re: [Mlt-devel] regarding irc channel for developers
Thanks brain for the quick reply :) I will compose my questions and mail it in the mailinglist. Regards, Bhuvan On Monday 19 March 2012 12:32 AM, Brian Matherly wrote: Is their an irc channel for developers? I think it is too much to expect but if their is a channel I would be delighted to join it and have a chat. I have so many questions to ask. There is a #mlt channel on irc.gimp.net. You will find that it is very quiet. The best way to get your questions answered is by reading the documentation: http://www.mltframework.org/twiki/bin/view/MLT/Documentation The second best way is by asking your questions right here in this mailing list. ~Brian -- This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure ___ Mlt-devel mailing list Mlt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mlt-devel
[Mlt-devel] decklink device enumerating cause a problem
Hi,
after recent patch for decklink producer and consumer i found a strange
behavior when first decklink working in play mode (process X) and second
decklink works in play-record-play mode (process Y).
Either some race condition or kind of deadlock happens on a driver
level. For some reason code line:
126 if ( m_deckLink-QueryInterface(
IID_IDeckLinkOutput, (void**) m_deckLinkOutput ) == S_OK )
start thread:
(gdb) bt
#0 0x003ec22e8217 in ioctl () from /lib64/libc.so.6
#1 0x7f1946d0743c in WaitForNotificationEvents () from
/usr/lib64/libDeckLinkAPI.so
#2 0x7f1946ca2633 in CDeckLinkOutput::DriverNotificationThread() ()
from /usr/lib64/libDeckLinkAPI.so
#3 0x7f1946ca2709 in
CDeckLinkOutput::DriverNotificationThreadFunction(void*) ()
from /usr/lib64/libDeckLinkAPI.so
#4 0x003ec2607d90 in start_thread () from /lib64/libpthread.so.0
#5 0x003ec22ef48d in clone () from /lib64/libc.so.6
that stall next line that release m_deckLinkOutput:
#0 0x003ec2609080 in pthread_join () from /lib64/libpthread.so.0
#1 0x7f8346a37b04 in CDeckLinkOutput::~CDeckLinkOutput() () from
/usr/lib64/libDeckLinkAPI.so
#2 0x7f8346a20cca in CDeckLink::~CDeckLink() () from
/usr/lib64/libDeckLinkAPI.so
#3 0x7f8346a2072f in CDeckLink::Release() () from
/usr/lib64/libDeckLinkAPI.so
#4 0x7f8346a48c01 in CDeckLink_v8_0::~CDeckLink_v8_0() () from
/usr/lib64/libDeckLinkAPI.so
#5 0x7f8346a48ea7 in CDeckLink_v8_0::Release() () from
/usr/lib64/libDeckLinkAPI.so
#6 0x7f834e1ee5cb in listDevices (properties=0x7f82ec016768,
this=0x7f82ec016760)
at consumer_decklink.cpp:142
First i revised all decklink pointers for any leak:
0001-care-about-NULL-pointers.patch
but that do not helps, so i checked mode iteration loop:
0002-make-sure-all-is-release-in-mode-iteration-loop.patch
but the cure was to disable device listing:
0003-make-device-listing-only-if-list_devices-property-se.patch
i dont know if it driver bug or some arch specific behaviour, but that
patches is an only way for me to provide stable decklink operation...
--
Maksym Veremeyenko
From 5ec0801c361a58628038c0ed9b334ae5086c36fc Mon Sep 17 00:00:00 2001
From: Maksym Veremeyenko ve...@m1stereo.tv
Date: Sat, 17 Mar 2012 15:53:40 +0100
Subject: [PATCH 1/3] care about NULL pointers
---
src/modules/decklink/consumer_decklink.cpp | 74 ++-
src/modules/decklink/producer_decklink.cpp | 49 +++
2 files changed, 67 insertions(+), 56 deletions(-)
diff --git a/src/modules/decklink/consumer_decklink.cpp b/src/modules/decklink/consumer_decklink.cpp
index 8645506..dceccdc 100644
--- a/src/modules/decklink/consumer_decklink.cpp
+++ b/src/modules/decklink/consumer_decklink.cpp
@@ -33,6 +33,8 @@
#include DeckLinkAPI.h
#endif
+#define SAFE_RELEASE(V) if (V) { V-Release(); V = NULL; }
+
static const unsigned PREROLL_MINIMUM = 3;
class DeckLinkConsumer
@@ -92,15 +94,19 @@ private:
public:
mlt_consumer getConsumer()
{ return m_consumer; }
-
+
+ DeckLinkConsumer()
+ {
+ m_deckLinkKeyer = NULL;
+ m_deckLinkOutput = NULL;
+ m_deckLink = NULL;
+ }
+
~DeckLinkConsumer()
{
- if ( m_deckLinkKeyer )
- m_deckLinkKeyer-Release();
- if ( m_deckLinkOutput )
- m_deckLinkOutput-Release();
- if ( m_deckLink )
- m_deckLink-Release();
+ SAFE_RELEASE( m_deckLinkKeyer );
+ SAFE_RELEASE( m_deckLinkOutput );
+ SAFE_RELEASE( m_deckLink );
}
bool listDevices( mlt_properties properties )
@@ -137,11 +143,11 @@ public:
free( key );
free( name );
}
- m_deckLinkOutput-Release();
+ SAFE_RELEASE( m_deckLinkOutput );
}
-m_deckLink-Release();
+SAFE_RELEASE( m_deckLink );
}
- decklinkIterator-Release();
+ SAFE_RELEASE( decklinkIterator );
mlt_properties_set_int( properties, devices, i );
mlt_log_verbose( NULL, [consumer decklink] devices = %d\n, i );
@@ -149,8 +155,7 @@ public:
}
catch ( const char *error )
{
- if ( decklinkIterator )
-decklinkIterator-Release();
+ SAFE_RELEASE( decklinkIterator );
mlt_log_error( getConsumer(), %s\n, error );
return false;
}
@@ -182,30 +187,32 @@ public:
return false;
}
#endif
-
+
// Connect to the Nth DeckLink instance
- do {
- if ( deckLinkIterator-Next( m_deckLink ) != S_OK )
- {
-mlt_log_error( getConsumer(), DeckLink card not found\n );
-deckLinkIterator-Release();
-return false;
- }
- } while ( ++i = card );
- deckLinkIterator-Release();
-
+ for ( i = 0; deckLinkIterator-Next( m_deckLink ) == S_OK ; i++)
+ {
+ if(i == card)
+break;
+ else
+SAFE_RELEASE( m_deckLink );
+ }
+ SAFE_RELEASE( deckLinkIterator );
+ if( !m_deckLink )
+ {
+ mlt_log_error( getConsumer(), DeckLink card not found\n );
+ return false;
+ }
+
// Obtain the audio/video output interface (IDeckLinkOutput)
if (
[Mlt-devel] scaler problem
Hi, converted image with melt has some strange artifact. # result broken image: /usr/local/mltframework.org/mlt/bin/melt -verbose -profile dv_pal -producer avformat:/home/studio/Videos/TheCore/hd/meta/test0002.mov.png in=1 out=1 -consumer avformat:/var/www/html/test_dv_pal.png vcodec=png f=image2 s=320x240 # result normal image: /usr/local/mltframework.org/mlt/bin/melt -verbose -profile square_pal -producer avformat:/home/studio/Videos/TheCore/hd/meta/test0002.mov.png in=1 out=1 -consumer avformat:/var/www/html/test_square_pal.png vcodec=png f=image2 s=320x240 source and resulting file could be downloaded from: http://downloads.m1stereo.tv/324c3cef50a05a05d2583456a1c6b21f/test0002.mov.png http://downloads.m1stereo.tv/324c3cef50a05a05d2583456a1c6b21f/test_dv_pal.png http://downloads.m1stereo.tv/324c3cef50a05a05d2583456a1c6b21f/test_square_pal.png mlt from git, ffmpeg from git, 64-bit linux... -- Maksym Veremeyenko -- This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure ___ Mlt-devel mailing list Mlt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mlt-devel
Re: [Mlt-devel] decklink device enumerating cause a problem
19.03.12 18:58, Dan Dennedy написав(ла): 2012/3/19 Maksym Veremeyenkove...@m1stereo.tv: Hi, after recent patch for decklink producer and consumer i found a strange behavior when first decklink working in play mode (process X) and second decklink works in play-record-play mode (process Y). Either some race condition or kind of deadlock happens on a driver level. For some reason code line: hmm, might be even worse when trying to use decklink producer and consumer in the same process! i used, but not simultaneously. it switch from avformat - decklink to decklink - avformat and seems working stable if device listing will be avoided... but the cure was to disable device listing: 0003-make-device-listing-only-if-list_devices-property-se.patch The only small problem I have with this is that it requires the app to start the consumer or fetch a frame from the producer in order to get the information. I can make a small improvement to this patch that establishes a property-changed event listener so that the enumeration occurs once when list_devices property is set, which is a little more convenient. as you prefer.. -- Maksym Veremeyenko -- This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure ___ Mlt-devel mailing list Mlt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mlt-devel
[Mlt-devel] Git: Initialize all decklink interface pointers and reset them upon release.
src/modules/decklink/consumer_decklink.cpp | 83 +++- src/modules/decklink/producer_decklink.cpp | 55 +++--- 2 files changed, 78 insertions(+), 60 deletions(-) New commits: commit edce14a421c21c1a190ade2d16d74753fa0c0511 Author: Maksym Veremeyenko ve...@m1stereo.tv Date: Mon Mar 19 20:53:32 2012 -0700 Initialize all decklink interface pointers and reset them upon release. Also, add a couple of missing releases. -- This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure ___ Mlt-devel mailing list Mlt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mlt-devel
Re: [Mlt-devel] build-melt.sh modifications
Of course, either location will work fine with wget or curl, but this is not about automation and build systems. This is about someone using a graphical web browser, downloading the script, and following some directions. The wiki inserts this HTTP header, which makes the browser download instead of display it: Content-disposition: inline;filename=build-melt.sh But I guess this is a minor point. If the user cannot figure out how to use Save As... or click Back and then right-click to download, then they should not be attempting to use this build script. I would tend to agree. A user can figure out how to save as. Otherwise, I wouldn't mind synchronizing the files with the Wiki manually. They shouldn't change that often. I am OK to use github since you plan to add other scripts for the build server. https://github.com/pez4brian/build-melt Just so I understand... Are you suggesting we keep all the scripts on github in my personal repository? It would work fine, but it seems a little arbitrary. I thought you could just create another repository at mltframework.org. Otherwise, we can create an mlt organization on github and keep the repo there. Either of those would make the most sense to me. as for v2 vs. v3 of build-melt.sh, I will take care of that in my fork and send you a pull request. It fixes some things on OS X for the nearly ready build agent. I resolved your comments in this commit: https://github.com/pez4brian/build-melt/commit/76cd4de9513947b84ec17de0b035881ad6f96477 ~BM -- This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure ___ Mlt-devel mailing list Mlt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mlt-devel
