commit fe5ffc294a4e183007326d36cb5594bfd409b16c
author François Degros <fdegros@chromium.org> Tue Mar 26 05:48:37 2024
committer Chromium LUCI CQ <chromium-scoped@luci-project-accounts.iam.gserviceaccount.com> Tue Mar 26 05:48:37 2024
tree 566d1c1007edbe386c3358a800a350f8af17fc78
parent cdff0aee8694de00a1f4418066074ee55e5d0abe

Revert "Mark WeakLinkNode sequence checking expensive"

This reverts commit 6f447d5040c0c6e619e51ccc53fe0d084c98e896.

Reason for revert: Crashes on ChromeOS as soon as I touch the
device's touchpad

2024-03-26T00:58:16.588106Z FATAL chrome[5433:5496]: [sequence_checker.cc(21)] Check failed: checker.CalledOnValidSequence(&bound_at). 
#0 0x56f06aa34da2 base::debug::CollectStackTrace()
#1 0x56f06aa19ba2 base::debug::StackTrace::StackTrace()
#2 0x56f06a934073 logging::LogMessage::Flush()
#3 0x56f063de6fcd logging::LogMessage::~LogMessage()
#4 0x56f06a92053c logging::(anonymous namespace)::DCheckLogMessage::~DCheckLogMessage()
#5 0x56f06a9205ae logging::(anonymous namespace)::DCheckLogMessage::~DCheckLogMessage()
#6 0x56f063dc2ba6 logging::CheckError::~CheckError()
#7 0x56f063de9d22 base::ScopedValidateSequenceChecker::ScopedValidateSequenceChecker()
#8 0x56f063de4706 base::ObserverList<>::begin()
#9 0x56f06bfa75eb ui::CursorController::SetCursorLocation()
#10 0x56f0646d9be5 ui::DrmCursor::SetCursorLocationLocked()
#11 0x56f0646da6bf ui::DrmCursor::MoveCursor()
#12 0x56f06bfd76e6 ui::GestureInterpreterLibevdevCros::OnGestureMove()


Original change's description:
> Mark WeakLinkNode sequence checking expensive
>
> This makes for-each-node sequence checking expensive (which seems
> redundant). As a compromise we do non-"expensive" sequence checking in
> ObserverList::begin(), which should provide the same level of protection
> unless iterators are passed between sequences, which would be one heck
> of a thing to try to do.
>
> This accounts for about 60% of sequence checking in a profile I did way
> back. I have not profiled to see how much sequence checking remains with
> the sequence checking moved to begin() nor do I know the average
> ObserverList size. Let's try it out.
>
> In the same profile (though I don't remember what I profiled) sequence
> checking accounted for 1.2% of cycles. Hopefully this explains some of
> the performance gap between a DCHECK and regular Canary build.
>
> Bug: 40241607
> Change-Id: Id80d3363771e05e6f38c1432ae66b4c352acf8b8
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/5319909
> Reviewed-by: François Degros <fdegros@chromium.org>
> Commit-Queue: Peter Boström <pbos@chromium.org>
> Reviewed-by: danakj <danakj@chromium.org>
> Reviewed-by: Vasiliy Telezhnikov <vasilyt@chromium.org>
> Reviewed-by: Colin Blundell <blundell@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#1277148}

Bug: 40241607
Change-Id: I8e11a7d796ba10a38453dff336f46e3aba04ab97
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/5394849
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Owners-Override: François Degros <fdegros@chromium.org>
Commit-Queue: François Degros <fdegros@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1278133}