<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style> benjiG: <u>hverkuil</u>: yes you can do that <br> <u>hverkuil</u>: I have add Patrice (sti DT maintainer) in copy for the DT patch hverkuil: thanks narmstrong: <u>hverkuil</u>: done, sent v5 hverkuil: <u>narmstrong</u>: thanks. I'll probably review it tomorrow and add my acks. narmstrong: thanks hverkuil: <u>narmstrong</u>: ping me if you haven't heard from me by Monday afternoon. milian: hey there. when I use gstreamer with v4l2 with a HD Webcam C525, I have a really long delay at startup for every application that uses this. strace shows me tons of ioctl of this form: https://paste.kde.org/pr7bkawcj (this is just the head, it continues like this for a long long time) <br> any idea what is going on there? <br> or how I could influence this to make startup faster? <br> note that I can reproduce this delay easily already with gst-launch-1.0 v4l2src ! xvimagesink <br> is this a gstreamer issue? or something with v4l2? <br> I notice that some other applications, presumably not using streamer, are not suffering from this, e.g. guvcviewer. so is this a setup question in the gstreamer pipeline? <br> hm vlc also doesn't show this issue, I'll try to ask the gstreamer people pinchartl: <u>milian</u>: it's a known gtsreamer issue with UVC devices <br> gstreamer tries all supported formats <br> and that's a costly operation with UVC cameras milian: can you point me at the issue report? I couldn't find anything on that <br> i.e. my googling failed me so far pinchartl: I can't sorry? I've discussed it maybe 5 years ago, I don't remember who with and where milian: ok hm <br> <u>pinchartl</u>: http://gstreamer-devel.966125.n4.nabble.com/Decreasing-startup-delay-with-v4lsrc-td4681453.html - thanks pH5: Is VIDIOC_SUBDEV_G_FMT supposed to return the field height (288 or 240 instead of the full frame height of 576 or 480) for V4L2_FIELD_ALTERNATE field order? hverkuil: <u>pH5</u>: I would expect so, since that's what VIDIOC_G_FMT will return as well. <br> the frame height makes no sense in combination with FIELD_ALT. pH5: How about VIDIOC_SUBDEV_ENUM_FRAME_SIZES ? <br> And V4L2_SEL_TGT_CROP ? hverkuil: crop is in frame coordinates, not field. <br> does frame_sizes make sense when used with FIELD_ALT? That usually means an HDTV/SDTV input, which typically does not implement ENUM_FRAME_SIZES. pH5: Ok, so in the spec "frame" is full frame as opposed to field, but "image" could be the field? hverkuil: I've never seen it used with interlaced formats, in fact. <br> yes. <br> No guarantee if the right terminology is always used :-) pH5: I want to make the i.MX6 CSI handle FIELD_ALT input (but output SEQ_BT or SEQ_TB depending on std) <br> with cropping and scaling hverkuil: Well, you're the first to do that. pH5: I hate interlacing :) hverkuil: Do you really need it? Handling interlaced formats is tricky, so I wouldn't implement it unless there is a good reason. <br> Main guideline for interlacing: everything uses frame coordinates, except for the actual buffers being transferred. So G/S/TRY_FMT all use half height for FIELD_ALTERNATE. <br> The vivid driver does it correctly (although it doesn't use subdevs of course). <br> v4l2-compliance also handles this correctly. pH5: Yes, I have a TPV5150 connected to i.MX6, and since commit 4f57d27be2a5 that produces FIELD_ALT output, and the CSI can only sync to full frames. <br> Excellent, thanks for clearing this up. hverkuil: Well, good luck! :-) pinchartl: <u>pH5</u>: from your question I assume that the documentation wasn't clear on the topic. it would be great if you had time to submit a documentation patch :-) pH5: <u>pinchartl</u>: already on it, if only to make sure I understood correctly from the patch feedback. eduardas_m: hello, where do I get (how do I install the Virtual Video Test Driver (vivid)) on my Ubuntu 14.04 system? I do not see it among the v4l-utils source hverkuil: <u>eduardas_m</u>: it's a kernel module, but not all distros build that driver. <br> Try (as root): modprobe vivid <br> If it is not found, then your distro doesn't build it and you have to compile the kernel yourself. eduardas_m: <u>hverkuil</u>: understood, thank you ***: benjiG has left pinchartl: <u>pH5</u>: thanks a lot!