Verified by manual revert on linux.

(of f8db70f30f0bdb3d6941000a3526d84ac63e820e)

Changes included in the roll:

$ git log e8f28818a..646e4293f --date=short --no-merges --format='%ad %ae %s'
2017-06-13 robertphillips Retract GrTexture*.h & GrRenderTarget*.h from other headers
2017-06-12 benjaminwagner [recipes] Fix internal_test tryjobs.
2017-06-13 rmistry Revert "Add GrPrimitiveType::kLinesAdjacency"
2017-06-13 mtklein ok: add 8888 and rp dsts
2017-06-12 csmartdalton Add GrPrimitiveType::kLinesAdjacency
2017-06-13 reed use pipeline for non-opt xfermodes
2017-06-13 robertphillips Move Flags from GrRenderTarget to GrTypesPriv.h & rename

Any ideas?
I went ahead and stopped the roller

This is Mike's CL, "2017-06-13 reed use pipeline for non-opt xfermodes".

Dropping sheriff label since revert landed.

I've reverted that CL in Skia and am about to re-start the roll into Chrome.

https://codereview.chromium.org/2936843003/ was supposed to suppress these failures.

That's weird.  Did we suppress them wrong?

I suspect something went wrong with converting them from being compared against the results of another html file rendering and being compared to a stored image.

So, yeah, but I don't know what went wrong.

Looks there are now images (png) for those tests. Perhaps I just re-mark those for needsmanualrebaseline, now that they are images and not html

This is now failing again after the roll, but this time on Windows:

https://uberchromegw.corp.google.com/i/chromium.webkit/builders/WebKit%20Win7%20%28dbg%29/builds/10064
https://chromium.googlesource.com/chromium/src/+/e635c3f42d2634a3335fea3671efeb95b3b898da

Please revert or update the suppressions for those tests.

Actually, nevermind, it seems fixed in latest builds. I was looking at an earlier one. Sorry for the noise!