[03:07] *** nst_ has quit IRC (Ping timeout: 272 seconds)
[06:13] *** LikWidChz has quit IRC (Ping timeout: 246 seconds)
[06:46] *** codyps has quit IRC (Quit: quit)
[07:34] *** bingbu has quit IRC (Ping timeout: 244 seconds)
[11:51] <hverkuil> paulk-leonov: ping
[12:43] <paulk-leonov> hverkuil, pong
[12:55] <hverkuil> paulk-leonov: quick question about the cedrus decoder: who processes information embedded in the bitstream such as the colorspace, transfer function, etc.? Is that userspace or the hardware?
[12:56] <hverkuil> I suspect that userspace has to parse this, right?
[12:56] <paulk-leonov> hverkuil, yes I think it's already part of the controls we have
[12:57] <hverkuil> it is?
[12:57] <hverkuil> I don't see it in include/media/mpeg2-ctrls.h
[12:59] <paulk-leonov> hverkuil, mhh I guess it depends on which bitstream elements you are referring to precisely
[13:00] <hverkuil> I have an MPEG bitstream that is decoded by a stateless decoder. A stateful decoder will extract such meta information and make it available to the driver. But what about a stateless decoder? I'm not sure where this information is stored (not an expert), so I don't know who is supposed to extract this information.
[13:05] <paulk-leonov> hverkuil, ah wait, I recall that we had an assumption on the YUV type at least
[13:07] <paulk-leonov> nope, chroma_format is in there
[13:10] <paulk-leonov> hverkuil, ok so there is indeed a transfer_characteristics field that we are not passing now
[13:15] <hverkuil> But it is userspace that parses this, and the only reason for passing it to the driver is so that the driver can correctly fill in the corresponding fields of v4l2_pix_format
[13:16] <hverkuil> (unless it is really fancy hardware that can do colorspace conversion, not impossible I guess)
[13:16] <paulk-leonov> hverkuil, yes it is userspace in any case
[13:17] <paulk-leonov> that's part of the things we need to improve in cedrus, right now it'll just assume NV12 for everything
[13:17] <hverkuil> transfer characteristics are independent of the pixelformat.
[13:18] <paulk-leonov> yes right, I'm mixing things up
[13:18] <paulk-leonov> hverkuil, in practice, it's likely that the player will parse this on its own and follow the info through to the display part independently from v4l2
[13:19] <hverkuil> yeah.
[13:19] <paulk-leonov> but since there are fields for the transfer, it would definitely make sense that they are filled up properly
[13:20] <hverkuil> I think we want to pass them along and let the driver fill in the v4l2_pix_format fields accordingly. And if you have a decoder that can do colorspace conversion, then it will need this information in any case.
[13:32] <paulk-leonov> definitely.
[13:34] <paulk-leonov> ezequielg, hverkuil: btw I have just updated master of https://github.com/bootlin/v4l2-request-test to include the out-of-tree header, use timestamps, make H264/H265 conditional and get rid of the DRM stuff that didn't make it upstream
[13:34] <paulk-leonov> should be easier to deal with
[13:34] <hverkuil> nice! Thank you.
[14:58] <syoung> shuah: there was a fix for the ir selftest: https://www.spinics.net/lists/linux-kselftest/msg05629.html now I need the same fix for another IR patch, see https://www.mail-archive.com/linux-media@vger.kernel.org/msg143235.html would be possible to have an Acked-B
[14:58] <syoung> y or have it merged via kselftree please? Thank you for your help.
[15:23] *** hverkuil has quit IRC (Quit: ZNC 1.7.1+deb2+b3 - https://znc.in)
[15:33] *** cybrNaut has quit IRC (Ping timeout: 244 seconds)
[17:04] *** benjiG has left 
[17:14] <ezequielg> paulk-leonov: i think the h264 headers needs to be imported.
[23:26] *** blinky42 has quit IRC (Ping timeout: 272 seconds)