Thu 22 May 2008 07:34:27 AM UTC, comment #10:
We can start a new bug for the issues in ffmpeg HEAD, if necessary.
|
Tue 20 May 2008 09:34:49 AM UTC, comment #9:
I confirm the bug is fixed in 0.8.3 branch.
In head, the NetStream-SquareTest.swf file changed a bit,
and ffmpeg breaks again there.
|
Thu 15 May 2008 06:02:26 AM UTC, comment #8:
And after a small amount of commenting things out...
(file #15667)
|
Thu 15 May 2008 05:55:22 AM UTC, comment #7:
There are two distinct issues causing this bug.
One: this is a small FLV file. When av_streamer() has let decoded the whole file (about ten frames), it returns, after setting the m_go status variable to FALSE. If m_go is false, video frames are no longer advanced and the movie is stuck on whatever data was available previous, in our case a bunch of zeroes because by the time av_streamer has processed the entire FLV file we haven't even started to display anything! The result is a black screen.
Two: refreshVideoFrame() relies on timings from the audio part of the decoder (i.e., audio_streamer sets m_current_timestamp) and it tests the presence of audio by the availability of a sound handler. But in our case there is no audio available from the FLV, so m_current_timestamp is never incremented. So, time stops, and the FLV never advances, even if the first issue above is resolved.
|
Wed 14 May 2008 07:00:36 AM UTC, comment #6:
The error message should be fixed in head, comes from parseNextFrame being unable to tell EOF from errors...
I'm assuming there's no tag to signal end of an FLV file, nor
a size header (which may be a wrong assumption, but I'm sure
gnash doesn't check for them).
The rectangle should be tilted 45 degrees, but should be yellow
with a smaller red square in it, the red square moving while
the movie plays.
The testcase now calls .pause initially, and calls .pause
again when you press space.
gstremer works much better as an example of how it should behave.
|
Wed 14 May 2008 06:54:34 AM UTC, comment #5:
I see a black rectangle tilted about 45 degrees. I see this error message:
20420:3069180816] 23:52:47 ERROR: FLVParser::parseNextFrame: can't read tag info (needed 12 bytes, only got 0)
|
Thu 10 Apr 2008 07:17:14 PM UTC, comment #4:
Hmmm. No way to edit posted comments, eh? It's not libavformat.h, it's avformat.h. At some point they moved it from .../ffmpeg/{avcodec.h,avformat.h} to .../avcodec/avcodec.h and .../avformat/avformat.h.
|
Thu 10 Apr 2008 07:01:44 PM UTC, comment #3:
Using svn from today, and cvs from today, I get this configure error:
ERROR: FFMPEG's libavcodec header is installed but not libavformat.
You can install FFMPEG from http://ffmpeg.mplayerhq.hu
or .deb users: apt-get install libavformat-dev
or reconfigure with --enable-media=gst
There is no avformat.h in ffmpeg, so I fear somebody is confused.
|
Wed 09 Apr 2008 05:10:46 PM UTC, comment #2:
Nelson, assigning to you as you're touching m4 macros... ok ?
Hong Yu said:
> We have used gnash0.8.2 to build with FFmpeg newer than Feb 18, 2008.
> And we have found that, very recent FFmpeg use expression/macro in
> defining 'LIBAVCODEC_VERSION' in avcodec.h, which caused Gnash0.8.2
> build to fail in defining 'FFMPEG_AUDIO2' and 'HAVE_SWSCALE_H' in
> gnashconfig.h. Gnash0.8.2 works fine when 'LIBAVCODEC_VERSION' is
> defined with string only (e.g., '51.50.1') in avcodec.h of FFmpeg. If
> you haven't handled such since gnash0.8.2, then very likely the
> potential problem is with current Gnash HEAD, too. Maybe you want to
> delay the bug report after you will have tried Gnash-cvs with newest
> FFmpeg? We are using FFmpeg source checked out on Feb 18, 2008.
|
Wed 09 Apr 2008 06:25:04 AM UTC, comment #1:
Hong Yu reports it to work with FFmpeg '51.50.1' and -O3.
Other experiences ?
|
Thu 28 Feb 2008 05:44:44 PM UTC, original submission:
That's all, with 0.8.2 branch and head ffmpeg fails to show the video.
Ffmpeg version is: Lavc1d.51.38.0
I get no ERROR line telling me why.
|