[PATCH] D46282: [AA] cfl-anders-aa with field sensitivity

[Chandler Carruth via Phabricator via llvm-commits]

chandlerc added subscribers: xbolva00, chandlerc.
chandlerc added a comment.

Folks, I want to be super, super clear here.

It is really not OK to take some previous patch that you didn't author,
take it over, and mail it out as a fresh patch. That is not how Open Source
works and that is not how LLVM works.

You need to attribute that patch to the author. You need to follow up with
the original authors and the original reviewers.

I'm particularly concerned because the previous approach had a *documented*
problem that both the author and George discussed in detail. This patch
should never have been committed (regardless of who says LGTM) until those
concerns have been discussed and addressed (or the folks with those
concerns found to be unavailable on completely unresponsive).

I'm calling this out here because this is super important to get right in
order to be an effective contributor to this (and most other) open source
projects.

Please *explicitly* and carefully document where this patch came from. Give
credit to Jia Chen who wrote the original patch. Discuss the delta (as you
are starting too, but please be more explicit), and make sure at least
George (the reviewer of the previous patch and owner of CFLAA) is happy,
and make *some* attempt to loop Jia Chen back in to cover any remaining
concerns there.

I *am* really happy to see folks picking up efforts that got stalled, and
helping out with them. That is a great pattern we should do more of in
LLVM. But we have to do it the right way, and make sure people are aware
and involved at the right points.


Repository:
  rL LLVM

https://reviews.llvm.org/D46282