Re: [libav-devel] [PATCH 3/8] lavc: Describe subtitle codecs as bitmap or text based.

2013-07-26 Thread wm4
On Fri, 26 Jul 2013 17:30:56 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 26/07/13 17:24, wm4 wrote: So everything was just shoehorned into something half-working. I understand that; but what does it have to do with the flag that the patch adds? The patch adds some information

Re: [libav-devel] [PATCH 3/8] lavc: Describe subtitle codecs as bitmap or text based.

2013-07-26 Thread wm4
On Fri, 26 Jul 2013 17:39:50 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 26/07/13 17:32, wm4 wrote: On Fri, 26 Jul 2013 17:30:56 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 26/07/13 17:24, wm4 wrote: So everything was just shoehorned into something half-working. I understand

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-16 Thread wm4
On Tue, 16 Jul 2013 11:05:22 +0200 Luca Barbato lu_z...@gentoo.org wrote: It maps part of API provided by AVFrame. --- libavcodec/avcodec.h | 48 ++ libavcodec/avpacket.c | 107 -- 2 files changed, 143 insertions(+), 12

Re: [libav-devel] [PATCH 3/3] api: Add public API to manipulate AVPacketList

2013-07-16 Thread wm4
On Tue, 16 Jul 2013 11:05:23 +0200 Luca Barbato lu_z...@gentoo.org wrote: +/** + * Remove the oldest AVPacket in the list and return it + * @param head List head. + * @param tail List tail. + * @param pkt Pointer to an AVPacket struct, its content will be + * overwritten. + *

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-15 Thread wm4
On Mon, 15 Jul 2013 15:07:16 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 07/14/2013 12:05 AM, wm4 wrote: Which perks does it actually have? Since even referencing an AVPacket requires memory allocation for the AVBufferRef, what do you win by not having to allocate the AVPacket? Our

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-13 Thread wm4
Should't the AVPacket documentation be extended about exact rules? For example, with AVFrame it's pretty clear that AVFrame can be changed by the owner of the AVFrame only. In particular, even the owner of an AVFrame reference has to call av_frame_make_writable() before changing the frame data

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-13 Thread wm4
PS: I challenge the need for these AVPacket inplace functions. You can't avoid memory allocation with it, because the buffer ref and the side-data require it anyway. So why is there a need for av_packet_ref() over av_packet_clone()? Maybe I'm overlooking something.

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-13 Thread wm4
On Sat, 13 Jul 2013 19:44:47 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 07/13/2013 07:25 PM, wm4 wrote: Should't the AVPacket documentation be extended about exact rules? For example, with AVFrame it's pretty clear that AVFrame can be changed by the owner of the AVFrame only

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-13 Thread wm4
On Sat, 13 Jul 2013 19:48:01 +0200 Luca Barbato lu_z...@gentoo.org wrote: On 07/13/2013 07:27 PM, wm4 wrote: PS: I challenge the need for these AVPacket inplace functions. You can't avoid memory allocation with it, because the buffer ref and the side-data require it anyway. So why

Re: [libav-devel] [PATCH 2/3] api: Add refcounted api to AVPacket

2013-07-13 Thread wm4
On Sat, 13 Jul 2013 21:56:22 +0200 Luca Barbato lu_z...@gentoo.org wrote: Personally, I think av_freep() is bad style (and probably even violates the C standard's struct aliasing rules) It does not violate the aliasing rules and it is just useful to prevent some common mistakes. As far

<    4   5   6   7   8   9