Re: [GIT PULL] RISC-V Patches for the 6.8 Merge Window, Part 3

[Palmer Dabbelt]

[Moving this back to linux-kernel@xxxxxxxxxxxxxxx, for some reason I changed my
alias to linux-kernel@xxxxxxxxxxxxxxxxxxx which doesn't actually exist.]

On Sat, 20 Jan 2024 09:42:08 PST (-0800), Linus Torvalds wrote:
> On Sat, 20 Jan 2024 at 06:52, Palmer Dabbelt <palmer@xxxxxxxxxxxx> wrote:
> > The last one is particularly clunky, as I've just squashed the fix into a merge
> > from your tree.
>
> _Please_ don't use back-merges to pre-merge things before sending me a
> pull request.
>
> Just let me know about the semantic conflict, and send your tree as it
> works for you. Then, as I merge your tree, I can take care of the
> actual conflicts.
>
> IOW, I really _really_ want developers to worry about *their* changes.
>
> Yes, linux-next is then there to give a heads-up about how these
> changes end up interacting, and you should be aware of it enough to
> send me a heads-up. But that's literally all - you should *not* then
> merge in other peoples changes into your tree.
>
> IOW, a conflict with other trees that got noticed in linux-next should
> just result in that note to me about the conflict once you send your
> tree to me.
>
> Not a random merge of a random commit that came in through another random tree.
>
> One result of that random merge is then your messy diffstat. Your pull
> request claims this:
>
>  443 files changed, 7415 insertions(+), 4420 deletions(-)
>
> but that's literally because you got that random noise from pulling in
> the state from the tty tree that had *nothing* to do with your RISC-V
> changes. The real diffstat if you don't have that last merge is
>
>  71 files changed, 2678 insertions(+), 215 deletions(-)
>
> and ends up looking much saner (ie it actually lists just arch/riscv/
> and the incidental changes around selftests etc that actually happened
> in your tree).

OK, sorry for that.

> So please send a new pull request where you just have *your* tree
> changes, not a random back-merge.

I think it's actually pretty simple for this one, I'm just backing up to 
the previous commit (as that nasty merge was at the tip of the tree).  
So I've set along 
https://lore.kernel.org/r/mhng-878cc879-7aa0-473f-91e4-8288072cdd4e@palmer-ri-x1c9/ 
, which doesn't have the merge and instead just calls out the diff in 
the comment.

>               Linus