On Sun, May 17, 2020, 5:41 PM Khem Raj <raj.k...@gmail.com> wrote: > > > On 5/17/20 6:22 AM, Adrian Bunk wrote: > > On Sat, May 16, 2020 at 10:24:04PM +0200, Andrey Zhizhikin wrote: > >> On Sat, May 16, 2020 at 10:10 PM Adrian Bunk <b...@stusta.de> wrote: > >>> > >>> On Sat, May 16, 2020 at 07:54:32PM +0000, Peter Kjellerstedt wrote: > >>>>> > >>>>> meta-openembedded/meta-python has a higher layer priority than > OE-core. > >>>>> > >>>>> Adding higher upstream versions of these recipes to a lower-priority > >>>>> layer in a stable series is a potential source for weird problems. > >>>> > >>>> I would assume they'd be removed from meta-python at the same time > >>>> they are added to meta. > >>> > >>> "at the same time" is complicated since OE-core has releases, > >>> but meta-openembedded is just a branch. > >>> > >>> I would also assume that not all users of stable series are updating > >>> meta-openembedded to the latest on the dunfell branch at the same > >>> time as OE-core, some users might end up updating one but never > >>> updating the other one. > >> > >> That I believe would be a terrible mistake when people opt-in to take > >> a layer, but never care about updating it... :( > > > > The typical Yocto user starts with whatever prehistoric Yocto release > > came with the BSP distribution for the reference hardware, and then > > develops a new product on top of that. > > > > I would not be surprised if someone will have the initial Yocto 2.7 > > release, but uses the latest from the corresponding meta-openembedded > > branch on top > > To be honest, we should not consider supporting such usecases IMO, > somewhere we have to draw a line, mixing layer releases is not something > upstream can even test those combos. >
I think it should be pretty clear that the product owner has the responsibility to update whatever BSP distribution they're using, and make that available to their users. At least IMO it was clear that users should maintain their distribution to the latest update of their stable branch, we can certainly put that on the docs or a wiki but I thought it was sort of given, since it simply makes sense, but perhaps making that assumption was a mistake. We can only do as much as bring updates after a certain release, wether its a case like this or its a CVE. We cannot control what some products do, yet we shouldn't let it stop us from doing changes that require updates in multiple layers if its a change that makes sense and most users will benefit from. Alejandro > > > > >> On the contrary, a quick run of "bitbake-layers show-overlayed" > >> exhibits the following on the [master] of both OE-Core and > > > > Thanks a lot for this. > > > >> meta-openembedded: > >> ================= > >> python3-cython: > >> meta-python 0.29.14 > >> meta 0.29.16 > >> python3-dbusmock: > >> meta-python 0.16.7 > >> meta 0.19 > >> python3-docutils: > >> meta-python 0.15.2 > >> meta 0.16 > >> python3-pyparsing: > >> meta-python 2.4.6 > >> meta 2.4.7 > >> ================= > > > > I'll take care of getting these removed from meta-openembedded. > > > >> Judging be versions, it looks like OE-Core recipes are maintained, but > >> in meta-openembedded they are left on the side... > > > > python3-docutils is ouch, since this problem is also in dunfell. > > > > On a more positive note 0.16 is already in OE-core in the initial > > Yocto 3.1 release, so removing it from meta-openembedded will only > > be like a 0.15.2 -> 0.16 upgrade for some users but cannot result > > in losing the recipe in weird layer combinations. > > > >> Regards, > >> Andrey. > > > > cu > > Adrian > > > > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#138394): https://lists.openembedded.org/g/openembedded-core/message/138394 Mute This Topic: https://lists.openembedded.org/mt/74214757/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-