Didn't look at the diff, but there's already code in the tree that requires python3 (eg subprocess.DEVNULL in compiler-rt/test/lit.common.cfg.py). So that part is fine.

We're looking into bumping libc++ in chromium, and this is a problem for us. We statically link libc++, and we still support 10.11. (We _just_ dropped support for 10.10 in our last release iirc.) What do you recommend as path forward?

Do we emit a warning if we see -gsplit-dwarf alone? Should we?

I'll revert this for now to unbreak the bot since figuring out a way forward will likely take a few hours.

This breaks hwasan tests on my bot: http://45.33.8.238/linux/37054/step_10.txt

Looks like this breaks tests on windows even after all the fix attempts: http://45.33.8.238/win/31260/step_11.txt

Hi, this broke check-clang everywhere as far as I can tell. Please run tests before committing. I've reverted this for now in 419ef38a50293c58078f830517f5e305068dbee6.

As far as I can tell, ld64 only accepts -undefined warning and -undefined suppress with -flat_namespace (which lld/MachO doesn't yet implement). That makes sense since with a flat namespace the undefineds can be found at dynamic link time, but with two-level lookup they can't since they containing framework isn't known (…right?).

Very nice!

It turned out to be UB in our code as far as I can tell, see https://bugs.chromium.org/p/angleproject/issues/detail?id=5500#c36 and the follow-on.

We bisected a test failure to this (https://bugs.chromium.org/p/angleproject/issues/detail?id=5500#c17). Can you expand a bit on what this patch means in practice? I'm guessing it makes UB in C++ code have bad effects more often? If so, what type of UB?

Looks good, thanks for the quick fix :)

Looks like this breaks tests on arm macs: http://45.33.8.238/macm1/1057/step_10.txt

Looks like this worked :)

I don't have a win machine at hand either, but let's land this and see what the win bot on http://45.33.8.238/ thinks about this. It should take at most 15 min to cycle.

Reverted in fe9976c02c09f105751f787ec998abeb3414a235 for now. If you could wait with relanding during working hours, that'd be super appreciated. Thanks!

In D87188#2470401, @lebedev.ri wrote:

In D87188#2470392, @thakis wrote:

Heads up: Breaks a test for us: https://bugs.chromium.org/p/chromium/issues/detail?id=1161542

(No reduced repro yet, might be UB, just fyi at this point.)

Thanks for headsup. For now i'll deal with the problem @nlopes pointed out above in a bit..

Looks like this breaks tests on mac: http://45.33.8.238/macm1/897/step_10.txt

Looks like this breaks tests on arm64-apple-macos: http://45.33.8.238/macm1/761/step_10.txt

Heads up: Breaks a test for us: https://bugs.chromium.org/p/chromium/issues/detail?id=1161542

...and two more in 1876a2914fe0bedf50f7be6a305f5bf35493e496. Sorry for the churn!

I reverted this (and a bunch of stuff that landed on top of it) in 7ad666798f12456d9e663e763e17e29007c3728d for now.

I verified it's due to this change. https://bugs.chromium.org/p/chromium/issues/detail?id=1161230#c15 has a stand-alone repro.

Heads-up: Looks like one of these changes, likely this one for CodeGen, made sanitizer output way larger: https://bugs.chromium.org/p/chromium/issues/detail?id=1161230

(nit: Consider checking in obviously-behavior-preserving changes like stripping namespace prefixes without pre-commit review, and only send out the behavior-changing bits.)

Thanks!

Thanks!

Looks like this breaks tests on windows: http://45.33.8.238/win/30322/step_8.txt

tweak -v help text

Thanks!

I just found this and realized I missed it for a very long time. I tried adding a test case, but while doing so I think this change doesn't actually change behavior: StringMapImpl::swap is only used by StringMap<T>::operator=, which guarantees that lhs and rhs have the same T and the underlying StringMapImpls the same ItemSize for that reason.

reverted in 00065d5cbd02b0f3fccb34881b58bcd0852b3970

Also, rnk's comment above wasn't addressed as far as I can tell.

This seems to break all tests on all platforms: http://45.33.8.238/linux/35854/step_7.txt

Actual bench.py / ministat before/after data: