I suspect kojii's patch, https://codereview.chromium.org/2865903002

kojii, can you take a look or revert if it's not possible to fix quickly?

Thanks, I'll take a look. Note, this is part of LayoutNG, which is not enabled on stable yet, so lowering the pri to 2.

Wondering if you knew, why is ClangToTLinuxASan bot catches what linux_chromium_asan_rel_ng doesn't? And what is the bot name if I want to run it? I can't confirm a fix without being able to try.

> Thanks, I'll take a look. Note, this is part of LayoutNG, which is not enabled on stable yet, so lowering the pri to 2.

Thanks. I don't care so much about the priority number, as long as it gets fixed :-) We shouldn't have broken code in the tree, regardless of whether it's enabled in the browser or not.

> Wondering if you knew, why is ClangToTLinuxASan bot catches what linux_chromium_asan_rel_ng doesn't?

It turns out to be because linux_chromium_asan_rel_ng builds with "dcheck_always_on=true" and ClangToTLinuxASan doesn't. Does your code do anything in a DCHECK that could affect the lifetime of some object?

> And what is the bot name if I want to run it? I can't confirm a fix without being able to try.

There is no trybot for ClangToTLinuxASan, but it's easy to reproduce locally:

$ gn gen out/lsan --args="is_asan=true is_lsan=true is_debug=false is_component_build=false use_goma=true"
$ ninja -C out/lsan -j1000 webkit_unit_tests
$ ASAN_OPTIONS="symbolize=1 detect_leaks=1" out/lsan/webkit_unit_tests --gtest_filter=NGLineBreakerTest.OverflowAtomicInline
==99138==ERROR: LeakSanitizer: detected memory leaks
...

I'm still seeing this. kojii, have you had a chance to take a look?

ping?

I opened the bug to check if the ping was for me. Looks like it was for kojii :-)

(protip: spell pings e.g. "kojii: ping", then fewer people have to click through)

kojii: Ping? Can you please fix or revert your change?

Sorry taking longer than expected. This is probably a combination of test runner and our bootstrap code, and I assumed we'll replace the bootstrap code with the real one soon, but it didn't happen as I expected. Probably a month or two more.

The best way for now I think is to do the same as  issue 721932 .

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/fc930616af698875d2e6a2a955326c7910f4059a

commit fc930616af698875d2e6a2a955326c7910f4059a
Author: kojii <kojii@chromium.org>
Date: Thu Jun 22 13:54:41 2017

Disable NGLineBreakerTest.OverflowAtomicInline under ASan

It leaks memory in CopyFragmentDataToLayoutBox. Disabling until
LayoutNGPaintFragments is ready. Same as  issue 721932 .

BUG= 732574 
R=hans@chromium.org, thakis@chromium.org
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_layout_ng

Review-Url: https://codereview.chromium.org/2952143002
Cr-Commit-Position: refs/heads/master@{#481519}

[modify] https://crrev.com/fc930616af698875d2e6a2a955326c7910f4059a/third_party/WebKit/Source/core/layout/ng/inline/ng_line_breaker_test.cc