On Wed, Sep 27, 2017 at 04:15:19PM +0200, Wolfgang Denk wrote:
This patch series reverts some patches that should never have been merged into mainline at all:
- They were applied in violation of the well established rule that ALL patches must be posted on the mailing list before, and given enough time for review by the community.
- Obviously, these patches have not been Acked-by: by anybody.
- They implement a solution that probably does not represent the majority of the developers who discussed this issue. They were applied cutting short the ongoing discussion, without giving a good, comprehensible reason. Alternative solutions were presented before, but this patch series ignores them.
U-Boot is a community project with an established and documented development process. No single person, not even a custodian or maintainer, shall have the power to push through any changes without or against the consent of the community. If we want to maintain these standards, we cannot accept this. We therefore have only one choice: revert these patches now, and let them undergo the necessary review process on the mailing list, allowing for the usual grace period of two weeks. Sorry.
Well, NAK. When you passed the "benevolent dictator" hat over, you passed it over. I value your input, and I don't lightly over-rule feedback. So, lets summarize things: - We need a modern device tree compiler. - Everyone needs a modern device tree compiler, because everyone should be running 'make tests'. - This is not the first, or second, or possibly even third time we've hit problems of people using different versions of dtc and seeing different results / problems. - The Linux Kernel has avoided this problem for forever by just bundling the minimum required parts of dtc in order to build things.
So, I've imported the infrastructure to keep things in sync, and imported upstream dtc.
Now, the best argument I've heard against this is "bundling is bad". I don't disagree, and the best way to fix this problem would be for distributions to start shipping a new enough version here that the kernel can drop it out, just like they did for linux-firmware recently. A whole lot more people care about the kernel than they do about U-Boot, and that's a more viable path forward. It would even make some kernel folks happy.