r327959 for  Issue 505314 

r328002 for natvis support in lld

Sanjay Patel committed r328119, which caused instcombine to assert in some cases. I went ahead and reverted it.

I think it's my turn, so I started a roll at r328575: https://chromium-review.googlesource.com/c/chromium/src/+/981314

This should have LLD /natvis support, the new pragma warning split, and a fix for infinite loops on generic lambdas with delayed template parsing (no crbug for that).

I'll have to go update the gn to deal with -Wno-ignored-pragma-optimize.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/9b87e595bffe9c67d7fd55feeb0c0d3a34b63bfa

commit 9b87e595bffe9c67d7fd55feeb0c0d3a34b63bfa
Author: Reid Kleckner <rnk@google.com>
Date: Tue Mar 27 11:33:45 2018

Roll clang 327688:328575.

Bug:  823655 
Change-Id: If6dfa9d775d3be37b72b4cf5f2fd9db89763819a
Reviewed-on: https://chromium-review.googlesource.com/981314
Reviewed-by: Hans Wennborg <hans@chromium.org>
Commit-Queue: Hans Wennborg <hans@chromium.org>
Cr-Commit-Position: refs/heads/master@{#546080}
[modify] https://crrev.com/9b87e595bffe9c67d7fd55feeb0c0d3a34b63bfa/build/config/compiler/BUILD.gn
[modify] https://crrev.com/9b87e595bffe9c67d7fd55feeb0c0d3a34b63bfa/tools/clang/scripts/update.py

We never resolved the Android ASan DomDistillerJsTest test failures. It looks like we don't have trybots for that configuration to use in the roll. If we have them on the clang waterfall, maybe we should add that configuration to the list of configs to test when rolling clang?

This appears to be sticking. Tracking bug for the next roll:  https://crbug.com/826683