1087537 - Linux e10s Mochitest 2 and 3 intermittently send a message to child process too late

Status: RESOLVED WORKSFORME

(Core :: DOM: Content Processes, defect, P5)

Description

[Andrew McCreight [:mccr8]]

Linux e10s M3 is intermittently failing to produce a log.  The one time I've managed to reproduce it on TBPL so far it was sending a bad message with this stack for the error:

###!!! [Child][OnMaybeDequeueOne] Error: Channel closing: too late to send/recv, messages will be lost
[Child 1870] ###!!! ASSERTION: Dropping message!: 'false', file /builds/slave/try-lx-d-000000000000000000000/build/dom/ipc/ContentChild.cpp, line 1660
#01: mozilla::dom::ContentChild::ProcessingError(mozilla::ipc::HasResultCodes::Result) [dom/ipc/ContentChild.cpp:1677]
#02: mozilla::dom::PContentChild::OnProcessingError(mozilla::ipc::HasResultCodes::Result) [obj-firefox/ipc/ipdl/PContentChild.cpp:5760]
#03: mozilla::ipc::MessageChannel::ReportConnectionError(char const*) const [ipc/glue/MessageChannel.cpp:1435]
#04: mozilla::ipc::MessageChannel::DequeueOne(IPC::Message*) [ipc/glue/MessageChannel.cpp:996]
#05: mozilla::ipc::MessageChannel::OnMaybeDequeueOne() [ipc/glue/MessageChannel.cpp:1019]
#06: RunnableMethod<mozilla::ipc::MessageChannel, bool (mozilla::ipc::MessageChannel::*)(), Tuple0>::Run() [ipc/chromium/src/base/tuple.h:383]
#07: mozilla::ipc::MessageChannel::DequeueTask::Run() [ipc/glue/MessageChannel.h:404]
#08: MessageLoop::RunTask(Task*) [ipc/chromium/src/base/message_loop.cc:362]
#09: MessageLoop::DeferOrRunPendingTask(MessageLoop::PendingTask const&) [ipc/chromium/src/base/message_loop.cc:372]
#10: MessageLoop::DoWork() [ipc/chromium/src/base/message_loop.cc:447]
#11: mozilla::ipc::DoWorkRunnable::Run() [ipc/glue/MessagePump.cpp:234]
#12: nsThread::ProcessNextEvent(bool, bool*) [xpcom/threads/nsThread.cpp:830]
#13: NS_ProcessNextEvent(nsIThread*, bool) [xpcom/glue/nsThreadUtils.cpp:265]
#14: mozilla::ipc::MessagePump::Run(base::MessagePump::Delegate*) [ipc/glue/MessagePump.cpp:100]
#15: MessageLoop::RunInternal() [ipc/chromium/src/base/message_loop.cc:233]
#16: MessageLoop::Run() [ipc/chromium/src/base/message_loop.cc:508]
#17: nsBaseAppShell::Run() [widget/xpwidgets/nsBaseAppShell.cpp:166]
#18: XRE_RunAppShell [toolkit/xre/nsEmbedFunctions.cpp:713]
#19: mozilla::ipc::MessagePumpForChildProcess::Run(base::MessagePump::Delegate*) [ipc/glue/MessagePump.cpp:272]
#20: MessageLoop::RunInternal() [ipc/chromium/src/base/message_loop.cc:233]

I think this indicates the parent process tries to send something too late. There's no easy way to tell what is sending it from this stack, so I'm not sure if this is a dupe of bug 1062472 or not.

Comment 1

[Andrew McCreight [:mccr8]]

I've now seen this 3 or so times, and every time there are precisely 54 such errors.

Comment 2

[Andrew McCreight [:mccr8]]

I've also seen this in one M2 run, where there's a similar failure stack over 100 times.

Comment 3

[Chris Peterson [:cpeterson]]

Intermittent e10s test failure

Comment 4

[Andrew McCreight [:mccr8]]

e10s content process leak checking has been working for a long time