[PATCH] D99305: [docs] Document our norms around reverts

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

In D99305#2651281 <https://reviews.llvm.org/D99305#2651281>, @rupprecht wrote:

> As a whole, I think this patch is going in the right direction -- we definitely need more concrete guidance on reverting patches.
>
> The decision tree here initially branches on whether the change is your own or someone else's change. Whoever authored/committed the patch certainly can come into play, but since the goal is to keep trunk in a good state, I think we should root the decision tree based on the content and impact of the change instead, e.g. deciding to revert a change based on the type of breakage (build failure, miscompile, performance degradation, compilation time/memory usage increase, etc.), how many platforms/users does it affect, what's the expected complexity/timeline of the fix, etc. That also emphasizes that the revert process is not personal, but just a matter of business keeping trunk green.

I personally feel this proposed reframing misses one of the really important parts of the proposed docs - that is, the interpersonal piece.  To me, probably the single most important piece in the whole doc is the bit about reverts generally being a favor to the commit author.  I really really hesitate to loose that.

I'm curious what others think, but I don't plan to take this suggestion unless the consensus from the community is that we should.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99305/new/

https://reviews.llvm.org/D99305