Hi vijay, I mean the adecoder src pad's caps in runtime did not match well with its template. It is just my guess, you'd better print out the runtime adecoder src0 caps. or you can refer the sourcecode in gstcaps.c: gst_caps_is_subset Best Regards Zhao Liang ________________________________ From: Vijay [mailto:vijay.linux.mail at gmail.com] Sent: Wednesday, April 16, 2008 3:28 PM To: Stefan Kost Cc: Zhao Liang-E3423C; gstreamer-embedded at lists.sourceforge.net Subject: Re: GStreamer on TI's embedded platform Zhao & Stefan, Thank you both, for your responses. I didnt' quite understand what you meant. As I understand caps negotiation, there should be no problem with the negotiation process if the adecoder sink pad (that feeds into the adecoder element) caps has an overlap with the filesrc src pad (that takes the data read from the file, by the filesrc element). And since the filesrc src pad is 'ANY', there should be an overlap always and therefore, there should be no problem. I'm not sure if that understanding is correct. Here's the entire adecoder caps values (obtained by using the command "gst-inspect-0.10 adecoder" on the command line) : Pad Templates: SRC template: 'src' Availability: Always Capabilities: audio/x-raw-int endianness: 1234 signed: true width: 16 depth: 16 rate: { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 } channels: [ 1, 2 ] SINK template: 'sink' Availability: Always Capabilities: audio/mpeg mpegversion: { 1, 2, 4 } layer: { 1, 2, 3 } audio/x-wma wmaversion: { 1, 2, 3 } rate: { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 } channels: [ 1, 2 ] Please let me know if my understanding is incorrect or if there could be some other problem. Thanks. Vijay On Mon, Apr 14, 2008 at 1:48 AM, Stefan Kost <ensonic at hora-obscura.de> wrote: Zhao Liang-E3423C schrieb: Vijay, I think you can open the log for more information, export GST_DEBUG=GST_CAPS:5 Seems actual adecoder src0 caps is not a subset of src template caps, maybe the src rate, channels, or depth is not suitable for template caps. This is a programming error in TIs elements. The quite likely set a value in the caps which are not in the template caps. Unfortunately you did not send those (just as "..."). But there they quite likely forgot e.g. rate or channels. Stefan *Best Regards Zhao Liang* ------------------------------------------------------------------------ *From:* Vijay [mailto:vijay.linux.mail at gmail.com] *Sent:* Thursday, April 10, 2008 9:38 PM *To:* Zhao Liang-E3423C *Cc:* gstreamer-embedded at lists.sourceforge.net *Subject:* Re: GStreamer on TI's embedded platform Thanks for your response, Zhao. I really appreciate it. I didn't get a chance to respond to your mail until now. I don't think I'm missing a mp3 demux element. Here's why: - This file has worked with every decoder I've tried, without any demuxing or parsing of container format (and I've tried a lot of decoders on a lot of systems! - all without any demuxing) - It has no ID3 or other tags (I removed all such MP3 headers and trailers in the file, myself) - It has only an elementary stream that contains only MP3 frames (each frame starting with "0xFF 0xFB"), all of the same length. That means, this file should not require a MP3 parser and any MP3 decoder should be able to decode the contents of the file directly. Plus, this pipeline comes from Texas Instruments, and they must have tested this before they put it on their website (hopefully)! Coming back to the problem I was facing: I was wrong about one thing. These are not two problems - it's probably one problem. It can't pause because it probably cannot negotiate caps. Here's what gst-inspect says about TI's adecoder plugin: Pad Templates: SRC template: 'src' Availability: Always Capabilities: audio/x-raw-int ... ... ... SINK template: 'sink' Availability: Always Capabilities: audio/mpeg mpegversion: { 1, 2, 4 } layer: { 1, 2, 3 } audio/x-wma wmaversion: { 1, 2, 3 } rate: { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 } channels: [ 1, 2 ] If I understand this correctly, it means that it takes mp3 or wma audio data as input and gives raw pcm samples as output. Since the filesrc element can output "any" data type, they should match and it should not be a problem. So I'm still stuck and I don't have a clue as to what the problem could be. Please, let me know if you have any idea about this problem. I'd appreciate any help. Thanks. Vijay On Tue, Apr 8, 2008 at 8:20 AM, Zhao Liang-E3423C <E3423C at motorola.com <mailto:E3423C at motorola.com>> wrote: Vijay, I think maybe you miss a mp3demuxe between filesrc and adecoder. and the command shall be gst-launch-0.10 filesrc location=$1 ! mp3demux ! adecoder Codec=3 ! osssink TI adecoder element may not accept the filesrc RAW caps. *Best Regards Zhao Liang * ------------------------------------------------------------------------ *From:* gstreamer-embedded-bounces at lists.sourceforge.net <mailto:gstreamer-embedded-bounces at lists.sourceforge.net> [mailto:gstreamer-embedded-bounces at lists.sourceforge.net <mailto:gstreamer-embedded-bounces at lists.sourceforge.net>] *On Behalf Of *Vijay *Sent:* Monday, April 07, 2008 8:32 PM *To:* gstreamer-embedded at lists.sourceforge.net <mailto:gstreamer-embedded at lists.sourceforge.net> *Subject:* GStreamer on TI's embedded platform Hi, I received TI's port of gstreamer to it's DaVinci processors from http://focus.ti.com/dsp/docs/dspsplash.tsp?contentId=3100 I've tried to run the example (scripts) provided by TI and I've faced what seem to be two separate issues. I've copied the stdout log below: <linux prompt>:/opt/system_files_gstreamer# ./test_MP3.sh MP3_file.mp3 Setting pipeline to PAUSED ... Pipeline is PREROLLING ... (gst-launch-0.10:1204): GStreamer-WARNING **: pad adecoder0:src returned caps which are not a real subset of its template caps (gst-launch-0.10:1204): GStreamer-CRITICAL **: gst_caps_get_structure: assertion `index < caps->structs->len' failed (gst-launch-0.10:1204): GStreamer-CRITICAL **: gst_structure_get_int: assertion `structure != NULL' failed [program hangs here] For most of you, who don't know about this script: All it does is after setting the gstreamer, plugin and library paths, it runs gst-launch with a pipeline, thus: gst-launch-0.10 filesrc location=$1 ! adecoder Codec=3 ! osssink The two issues I'm facing: (a) The *last* element in the gstreamer pipeline does not reply with a message to the app saying it has paused (This is ascertained with debug prints that I inserted in gst-launch.c. It's possible that for some reason, the element does not pause. I've faced this same issue with pipelines which have no decode/render elements as well.) Actually, I'm not sure if (a) it doesn't pause because it doesn't finish preroll or (b) it says it hasn't finished preroll because it doesn't send a pause signal! (I'm not sure which is the cause and which is the effect). (b) The critical errors printed (seen above. I guess these are caused because of TI's mp3 decoder element plugin.) Could the adecoder capabilities be incompatible? Seems unlikely, since TI would have tested this first. Would anyone know what the issue might be? Has anyone seen a similar issue with gstreamer? I'd greatly appreciate any help. Thanks. Vijay -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.freedesktop.org/archives/gstreamer-embedded/attachments/20080416/df5f4e5f/attachment.htm>
