Deprecation policy for old BSPs? #617
Replies: 5 comments 9 replies
|
I think deprecating old boards sounds like a good idea |
|
I like the idea as well. We need a way to park them and perhaps a file called |
|
Totally agree. This will help to keep the layer clean, and share "clean" code between well supported boards. |
|
Hi all, I have created an initial PR for deprecation warnings in the layer: #625 Right now I have only included What do you all think? |
|
Thanks Khem Raj (@kraj) for merging my initial PR. As mentioned above, it warns about the following
What do you all think about adding to the deprecation list for 6.1? Do we need to keep the vendor variants for the K1 boards? Are there other BSPs you think we should deprecate? |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I've been looking at the list of MACHINE options in the layer for a while, and (especially after creating the table listing them) I think we ought to have a policy for removing older and/or under-supported platforms. Here's a short list of examples:
The StarFive VisionFive (V1) does not even appear to be listed on their product list anymore, and it was three pages in a Google query for "starfive visionfive" before I saw a link dedicated to it instead of the VisionFive 2.
From an availability perspective: the BeagleV, which had a very limited run and is unlikely to get any future support.
Others like the Star64 may also be considered for removal over time simply because of the legacy kernel versions they use, although those should probably lower-priority than boards that aren't available at all (and someone may still be working on updating them).
Finally, I think it's worth considering whether or not we should have a removal timeline for boards which we currently support with vendor kernels, but are also seeing active development upstream and in this layer with mainline repositories. The OrangePi RV2 and other K1 boards are a good example - I'm not suggesting that we have to drop vendor support immediately, but from a maintenance perspective it may not be worth keeping both versions.
The greatest motivation for this, in my opinion, is making sure that we can reliably automate testing for all of the boards while focusing primarily on the ones that the community will have access to.
Thoughts?
All reactions