[gst-devel] [gst-cvs] wtay gst-plugins-good: gst-plugins-good/gst-plugins-good/gst/rtp/

[Peter Kjellerstedt]

> -----Original Message-----
> From: gstreamer-cvs-bounces at lists.sourceforge.net 
> [mailto:gstreamer-cvs-bounces at lists.sourceforge.net] On 
> Behalf Of wtay at kemper.freedesktop.org
> Sent: den 11 december 2007 17:32
> To: gstreamer-cvs at lists.sourceforge.net
> Subject: [gst-cvs] wtay gst-plugins-good: 
> gst-plugins-good/gst-plugins-good/gst/rtp/
> 
> CVS Root:       /cvs/gstreamer
> Module:         gst-plugins-good
> Changes by:     wtay
> Date:           Tue Dec 11 2007  16:32:04 UTC
> 
> Log message:
> Patch by: Wai-Ming Ho <webregbox at yahoo dot co dot uk>
> * gst/rtp/gstrtph264pay.c: (gst_rtp_h264_pay_init),
> (gst_rtp_h264_pay_finalize), (gst_rtp_h264_pay_setcaps),
> (next_start_code), (is_nal_equal), (gst_rtp_h264_pay_decode_nal),
> (encode_base64), (gst_rtp_h264_pay_parse_sps_pps),
> (gst_rtp_h264_pay_handle_buffer):
> * gst/rtp/gstrtph264pay.h:
> Use higher performance start-code searching.
> Parse NALs and store SPS, PPS and profile in the caps so that they can
> be used in the SDP. Fixes #502814.
> 
> Modified files:
>     .               : ChangeLog
>     gst/rtp         : gstrtph264pay.c gstrtph264pay.h
> 
> Links:
>
http://freedesktop.org/cgi-bin/viewcvs.cgi/gstreamer/gst-plugins-good/Ch
angeLog.diff?r1=1.3212&r2=1.3213
>
http://freedesktop.org/cgi-bin/viewcvs.cgi/gstreamer/gst-plugins-good/gs
t/rtp/gstrtph264pay.c.diff?r1=1.3&r2=1.4
>
http://freedesktop.org/cgi-bin/viewcvs.cgi/gstreamer/gst-plugins-good/gs
t/rtp/gstrtph264pay.h.diff?r1=1.1&r2=1.2

* In gst_rtp_h264_pay_finalize(): 
  Why check for NULL before calling g_free()?
* In encode_base64():
  Why implement yet another base64-encoder? Now there is one
  in the RTSP-code and one here. Wouldn't it make sense to have
  one implementation for all of GStreamer? And if glib >= 2.12
  then I guess you could use g_base64_encode(), and not need
  to have any at all.
* Why whould is_nal_equal() be faster than memcmp()?
  On what architectures?
  Oh, and the comment preceding it mentions memcpy rather 
  than memcmp()...
* Why not make gst_rtp_h264_pay_decode_nal() return a gboolean
  denoting the updated status, rather than making it take a
  gboolean return parameter?

//Peter