Hi Martin, On 1/9/24 11:17, Martin Tůma wrote: > Hi, > If a driver implements VIDIOC_ENUM_FRAMESIZES for a output device, qv4l2 > crashes in general-tab.cpp:2169 due to m_frameSize being 0. As all other > usages of m_frameSize in the GeneralTab::updateFrameSize() function are > guarded by "NULL checks" an obvious fix would be to guard the "addItem > while cycle" as well. But maybe a better solution would be to add the > frame size combobox to output devices as well. Or are the output devices > not supposed to have frame sizes enumeration? You do not expect to see it for output devices. Those will typically use VIDIOC_ENUMSTD or VIDIOC_ENUM_DV_TIMINGS. VIDIOC_ENUM_FRAMESIZES is something that only makes sense for output devices if they do not support ENUMSTD or ENUM_DV_TIMINGS, i.e. it has some very odd output hardware. In particular vivid doesn't support this for the emulated video output. I don't think v4l2-compliance has a check for this. I think it should at least issue a warning if the video output devices supports ENUM_FRAMESIZES and also ENUMSTD or ENUM_DV_TIMINGS. In any case, qv4l2 shouldn't crash, so an initial fix should be to check for m_frameSize being 0. A patch is welcome. Which driver is this? I assume it is an out-of-tree driver, since I am not aware of any mainline drivers that do this. Regards, Hans
