On Tue, 2026-06-16 at 11:48 +0200, Michael Opdenacker wrote: > > On 6/16/26 8:45 AM, Alex Kiernan via lists.openembedded.org wrote: > > On Mon, Feb 5, 2024 at 4:54 PM Richard Purdie > > <[email protected]> wrote: > > > In talking with some others about layer setup and configuration, we > > > were talking about some of the challenges and the general confusion > > > that there is, particularly around priorities. > > > > > > We asked the question "when do you actually need to use priorities?". > > > The answer seemed to be not that often in reality. > > > > > > That leads to the question, could we just drop it? > > > > > > Amongst a small group of us, that had a surprising amount of support, > > > enough that I've decided to write the idea down and put it on the list. > > > > > > Yes, removing it might break a few things however it could be a big > > > usability and understandably win. The people I've talked to so far > > > would happily fix any fallout for that win. > > > > > > Just for completeness, I want to mention that the history of this comes > > > from pre layer days when we had "collections" as a poor version of > > > layers. I believe layers, BBMASK, bbappends and the fact we have much > > > few recipes with multiple versions all point to this as being obsolete > > > technology we no longer need. > > > > > > Thoughts? > > > > > I just got bitten by this again... I have a need of an older > > linux-libc-headers for a particular machine, had wired everything up > > so it built as I wanted, but then when I switched back to a sensible > > machine, was surprised when I still got old headers, because of > > BBFILE_PRIORITY. > > > > Still a +1 from me to get rid of it! > > I vote for this too. > This sounds risk to rely on priorities to override things. > Priorities could change and have unexpected ripple effects.
I'm not sure how much was recorded at the time but I got into huge amounts of trouble over this and some people are still holding a grudge against me over it, even now. For the record, the OE TSC did discuss it, got excited about being able to simplify things and missed a line of code linking things together which was rather critical. When we realised that, we basically dropped the proposal as the basis of the reasoning was flawed. We do have an aim of trying to revisit this and work out what we could do to try and simplify layers, ideally around now in the release cycle. Unfortunately I keep getting pulled in different directions and after getting burnt so badly last time, I'm reluctant to say much without careful thought. There was some discussion on the weekly tech call a couple of weeks ago but it was just discussion and there is no actual proposal yet. So yes, I would love to do something but there are also people who are unwilling to change what they're doing and potentially lose some configuration options. Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#2401): https://lists.openembedded.org/g/openembedded-architecture/message/2401 Mute This Topic: https://lists.openembedded.org/mt/104180197/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-architecture/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
