Hi Greg, Krzysztof, Le mardi 13 mai 2025 à 07:04 +0200, Greg KH a écrit : > On Mon, May 12, 2025 at 11:03:41PM +0200, Krzysztof Opasiak wrote: > > On 12.05.2025 12:43, Krzysztof Opasiak wrote: > > > On 12.05.2025 12:38, Greg KH wrote: > > > > On Mon, May 12, 2025 at 12:19:07PM +0200, Krzysztof Opasiak wrote: > > > > > Hi Greg, > > > > > > > > > > On 4.12.2022 09:29, Greg KH wrote: > > > > > > On Sat, Dec 03, 2022 at 11:26:14PM +0200, Laurent Pinchart wrote: > > > > > > > Hi Michael, > > > > > > > > > > > > > > On Sat, Sep 10, 2022 at 12:13:31AM +0200, Michael Grzeschik wrote: > > [...] > > > > > > > > > > Given that I'd like to suggest that it seems to actually make sense to > > > > > revert this unless there are some ideas how to fix it. > > > > > > > > Sorry about this, can you submit a patch series that reverts the > > > > offending commits? As it was years ago, I don't exactly know what you > > > > are referring to anymore. > > > > > > > > > > Sure! Will do. > > > > > > > Would you prefer to have a set of actual reverts related to this: > > > > da692963df4e Revert "usb: gadget: uvc: add v4l2 enumeration api calls" > > bca75df69aaf Revert "usb: gadget: uvc: add v4l2 try_format api call" > > e56c767a6d3c Revert "usb: gadget: uvc: also use try_format in set_format" > > 20f275b86960 Revert "usb: gadget: uvc: fix try format returns on > > uncompressed formats" > > 059d98f60c21 Revert "usb: gadget: uvc: Fix ERR_PTR dereference in > > uvc_v4l2.c" > > e6fd9b67414c Revert "usb: gadget: webcam: Make g_webcam loadable again" > > > > but have a negative consequence that the series isn't really bisectable from > > functional perspective. For example commit e6fd9b67414c breaks g_uvc until > > we apply da692963df4e so the series would have to go in as a whole. > > > > Or you would prefer a single commit that technically isn't a revert but it > > just "undoes" the negative consequences of "usb: gadget: uvc: add v4l2 > > enumeration api calls" (kind of a squash of all commits above)? > > Ideally we can bisect at all places in the tree, so it's odd that > reverting patches would cause problems as when adding them all should > have been ok for every commit, right? > > But if there are merge issues, or other problems, then yes, maybe just > one big one is needed, your choice. Won't a plain revert break userspace like GStreamer that have depended on that patch for years ? In such a delicate case, wouldn't it be less damageable to introduce workaround, like alias ? This is one personal script against numerous users of a generic framework implementation. I believe due to the delay, you are facing an unusual ABI breakage, which requires a more delicate handling. regards, Nicolas > > thanks, > > greg k-h
