hverkuil_: Would you have time today to discuss IOCTL argument matters? ndufresne: hi again, what about this patch? I disabled dmabuf if an emulated pixel format is detected. -> http://paste.debian.net/1019792/ Whoopie, except for the return else return, this patch seems to make sense ndufresne: what's your preferred coding style? Whoopie, Would filling a bug and attaching the patch (create it with git format-path) to bugs.gnome.org Whoopie, the git repo have a commit hook that will tell you ;-P but just return faster the if in this case no need for the else ndufresne: I attached it to https://bugzilla.gnome.org/show_bug.cgi?id=795097 I placed it now in front of the ioctl to save time. no need for it when it's an emulated format. Whoopie, thanks, all I need to check now, is that gst_v4l2_object_is_dmabuf_supported() is not called because S_FMT hmm, what does that mean? ;-) looks fine, it's called by _setup_pool() which is called right after _save_format() Whoopie, means that I had to check if v4l2object->fmtdesc when this function is called ok I guess a null check on fmtdesc would make it extra safer, in case someone starts using this function in another context ... ndufresne: but that means that if fmtdesc was NULL, DMABuf would not be disabled, right? that's why I'd make it an assert instead of a null check ah wait, I know Whoopie, I'd move the check inside setup_pool this way the generic test can be used whenever needed, and we are not enabling dmabuf when we create a pool or not bother at all, I realize it's internal Whoopie, yeah, leave your patch like this, I think it's fine sorry for the noise thanks for the feedback, I really appreciate it! well, thanks for taking the time to write this patch this is always a challenge for me. Whoopie, so with that, + the env, do you get cheese working ? (or what's the next step ?)