[+mgersh]:  Mind investigating?

Do you have links to bot runs where this happened?

The DnsConfigService conversion to TaskScheduler was only a few months ago, I wouldn't be surprised to still find some bugs. It looks like it's failing in initialization, not shutdown.

No. I only have these screenshots from tests that run *after* the failing test on the same bot. I don't know what it was that ran before interactive_ui_tests on this particular swarming bot. I've seen this same error message on many runs.

Maybe someone from infra could help figure out what is popping the dialog. +maruel, who added some logging recently that might help.

Now that CHROME_HEADLESS=1 is properly being set (see issue 790985), we're seeing this crash in failing browser_tests; see, for example, https://chromium-swarm.appspot.com/task?id=3a2d610a75456610&refresh=10&show_raw=1:

[ RUN      ] ServiceProcessControlBrowserTest.DieOnDisconnect
...
[4372:6280:1201/161614.652:FATAL:post_task.cc(72)] Check failed: TaskScheduler::GetInstance(). Ref. Prerequisite section of post_task.h.

Hint: if this is in a unit test, you're likely merely missing a base::test::ScopedTaskEnvironment member in your fixture.

Backtrace:
	base::debug::StackTrace::StackTrace [0x02BBD2B0+32]
	base::debug::StackTrace::StackTrace [0x02B7D2ED+13]
	logging::LogMessage::~LogMessage [0x02B2801E+78]
	base::PostDelayedTaskWithTraits [0x02B52F01+97]
	base::PostTaskWithTraits [0x02B53081+65]
	base::CreateCOMSTATaskRunnerWithTraits [0x02B53455+277]
	base::internal::PostTaskAndReplyImpl::PostTaskAndReply [0x02BC6D51+625]
	base::PostTaskWithTraitsAndReply [0x02B53011+97]
	net::SerialWorker::WorkNow [0x02E87709+329]
	net::internal::DnsConfigServiceWin::ReadNow [0x02DF5A17+23]
	net::DnsConfigService::WatchConfig [0x02D771CB+217]
	net::NetworkChangeNotifierWin::DnsConfigServiceThread::Init [0x02D76949+169]
	base::Thread::ThreadMain [0x02B6EBD4+548]
	base::PlatformThread::GetCurrentThreadPriority [0x02B4B1D3+467]
	BaseThreadInitThunk [0x75D3337A+18]
	RtlInitializeExceptionChain [0x77C692B2+99]
	RtlInitializeExceptionChain [0x77C69285+54]

Received fatal exception EXCEPTION_BREAKPOINT
Backtrace:
	base::debug::BreakDebugger [0x02BC656C+12]
	?Run@?$Invoker@U?$BindState@P6AXPBDHV?$BasicStringPiece@V?$basic_string@DU?$char_traits@D@std@@V?$allocator@D@2@@std@@@base@@1@Z$$V@internal@base@@$$A6AXPBDHV?$BasicStringPiece@V?$basic_string@DU?$char_traits@D@std@@V?$allocator@D@2@@std@@@3@1@Z@internal@ [0x0353081F+31]
	logging::LogMessage::~LogMessage [0x02B283DE+1038]
	base::PostDelayedTaskWithTraits [0x02B52F01+97]
	base::PostTaskWithTraits [0x02B53081+65]
	base::CreateCOMSTATaskRunnerWithTraits [0x02B53455+277]
	base::internal::PostTaskAndReplyImpl::PostTaskAndReply [0x02BC6D51+625]
	base::PostTaskWithTraitsAndReply [0x02B53011+97]
	net::SerialWorker::WorkNow [0x02E87709+329]
	net::internal::DnsConfigServiceWin::ReadNow [0x02DF5A17+23]
	net::DnsConfigService::WatchConfig [0x02D771CB+217]
	net::NetworkChangeNotifierWin::DnsConfigServiceThread::Init [0x02D76949+169]
	base::Thread::ThreadMain [0x02B6EBD4+548]
	base::PlatformThread::GetCurrentThreadPriority [0x02B4B1D3+467]
	BaseThreadInitThunk [0x75D3337A+18]
	RtlInitializeExceptionChain [0x77C692B2+99]
	RtlInitializeExceptionChain [0x77C69285+54]
../../chrome/browser/service_process/service_process_control_browsertest.cc(126): error: Failed
../../chrome/browser/service_process/service_process_control_browsertest.cc(314): error: Value of: ServiceProcessControl::GetInstance()->IsConnected()
  Actual: false
Expected: true
[264:7424:1201/161616.580:INFO:chrome_cryptauth_service.cc(222)] Profile is not authenticated yet; waiting before starting CryptAuth managers.
[264:7424:1201/161616.890:WARNING:url_request_context_getter.cc(43)] URLRequestContextGetter leaking due to no owning thread.
[  FAILED  ] ServiceProcessControlBrowserTest.DieOnDisconnect, where TypeParam =  and GetParam() =  (2635 ms)

Without CHROME_HEADLESS=1, the dialog in the initial report is shown, which was likely leading to test timeouts. Now we're at least seeing the error where it happens. This is leading to flaky browser_tests, so please fix as soon as possible. I have bumped this to P1 since it should now be quite actionable.

It sounds like this is a problem with initialization order in a specific set of tests, which already has a bug filed. The DnsConfigService initialization just happens to be something that needs the TaskScheduler early on. I don't think I'm the right person to debug this, but let me know if you see it anywhere other than ServiceProcessControl tests.

The crash stacks for the initial reports in issue 517420 have nothing to do with those in comment 4 here. I don't think it's helpful to lump unrelated problems into the same bug.

I just poked around in codesearch for a few minutes, and it seems pretty clear that the problem is in chrome/service/service_process.cc:

bool ServiceProcess::Initialize(base::MessageLoopForUI* message_loop,
                                const base::CommandLine& command_line,
                                ServiceProcessState* state) {
...
  network_change_notifier_.reset(net::NetworkChangeNotifier::Create());
...
  base::TaskScheduler::Create("CloudPrintServiceProcess");
...
}

those two lines are out-of-order. Francois, could you take this?

Oh, and if I'm reading it right, this is a product problem, not a test problem. That is, it's not initialization order in a specific set of tests,

I can believe it's unrelated to the Mac flakiness, I didn't look. But it's definitely the cause of the Windows flakiness in comments 18-20 on the other bug, I did check those links, and I don't think it's helpful to not inform the people working on a bug that you've decided to separate it out. I'll go comment there.

Good point. Thanks for updating the other bug. Cheers.

Upon re-reading comment 6, I think I came across as more gruff than is appropriate. My apologies.

https://chromium-review.googlesource.com/c/chromium/src/+/810925

 Issue 791791  has been merged into this issue.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/1d7a91284b1ea021daaeeea39c8f6a242782f670

commit 1d7a91284b1ea021daaeeea39c8f6a242782f670
Author: Francois Doray <fdoray@chromium.org>
Date: Thu Dec 07 00:01:50 2017

Initialize NetworkChangeNotifier after TaskScheduler in service process.

The NetworkChangeNotifier must be created after TaskScheduler because it
posts tasks to it.

Bug:  789442 
Change-Id: Ie6066064a5f840666160bac76d95ad16d3363708
Reviewed-on: https://chromium-review.googlesource.com/810925
Reviewed-by: Vitaly Buka <vitalybuka@chromium.org>
Commit-Queue: François Doray <fdoray@chromium.org>
Cr-Commit-Position: refs/heads/master@{#522262}
[modify] https://crrev.com/1d7a91284b1ea021daaeeea39c8f6a242782f670/chrome/service/service_process.cc

Thanks for the fix! Would you also please re-enable the tests that were disabled for this bug? https://bugs.chromium.org/p/chromium/issues/detail?id=517420#c18 and https://bugs.chromium.org/p/chromium/issues/detail?id=517420#c20 might both be okay now.

CL to re-enabled the tests waiting for review https://chromium-review.googlesource.com/c/chromium/src/+/817574

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/29c78ed184f0b7c2a3e71462450030d72ce0bc70

commit 29c78ed184f0b7c2a3e71462450030d72ce0bc70
Author: Francois Doray <fdoray@chromium.org>
Date: Fri Dec 08 23:23:38 2017

Re-enable ServiceProcessControlBrowserTest on Windows.

This reverts commits
d70e91da314d7d5b28bcfe97007933dc0aec9dfd and
448f299e4fb9825df68bad77b8f2352b288f5a63.

Issue on Windows was fixed by
https://chromium-review.googlesource.com/810925.

Issue on Mac seems to have a different root cause.
We should consider re-enabling the tests, which were
disabled in 2015, and see if the issue is still there.

Bug:  789442 , 517420
Change-Id: Id4ce64b5ec1ffa6d97c25b7967a92ac0ad293979
Reviewed-on: https://chromium-review.googlesource.com/817574
Reviewed-by: Vitaly Buka <vitalybuka@chromium.org>
Commit-Queue: Vitaly Buka <vitalybuka@chromium.org>
Cr-Commit-Position: refs/heads/master@{#522915}
[modify] https://crrev.com/29c78ed184f0b7c2a3e71462450030d72ce0bc70/chrome/browser/service_process/service_process_control_browsertest.cc

Thanks!