GitLab

[Reland] - This is effectively a reland, first patchset is unchanged
https://crrev.com/74f12a609316ba81 so compare to that for changes. That
iteration provided some interesting data but used a triggering time of 5
seconds. I've since learned that there's a myriad of known edge-cases in
which events might have delays of 1-20 seconds (!). In our case, we're
specifically looking for issues at the upper end of the UMA metric (60+
seconds) so this caused some unrelated noise. This CL relands with a
higher threshold and some minor changes to the crash keys.

Add a DumpWithoutCrash to KeyPress events where the timestamp of the OS
differs from the current time by more than 60 seconds. We see in UMA
that we often see events in the 95% percentile reaching many seconds of
difference. Hopefully the crash stacks will reveal something about
what's causing this.

A previous iteration of this investigation in
https://crrev.com/c/2111782 showed that we sometimes redispatch events
that have been reinjected back from the renderer. This should never
happen. This CL adds crash keys to dump some of the related app state
to help explain how this happens.

However, not all such stack traces showed reinjection so that can't be
the cause of all these bad timestamps. This CL also adds some timing
information to the existing crash key in NSApp:sendEvent which might
reveal something.

Bug: 1039833
Change-Id: If49b7b8a918f880ed00a7adf219e589ed66daf6e
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2210867
Commit-Queue: David Bokan <bokan@chromium.org>
Reviewed-by: Avi Drissman <avi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#771059}