GitLab

ResetIPC() is called by both Cleanup() and ProcessDied(), where the
latter may be called immediately before the RPHI is reused for a new
render process.

Some IPC-related cleanup tasks were only being done in Cleanup(). This
means it was possible for some tasks scheduled on behalf of the previous
process to run after a new process was started by the host. Namely
UI-thread interface binding tasks -- limited in lifetime by
instance_weak_factory_ WeakPtrs -- could run despite ProcessDied() being
invoked after their scheduling.

This CL resets instance_weak_factory_ in ResetIPC() rather than only in
Cleanup(). It also moves |io_thread_host_| cleanup to ResetIPC() for
good measure, since this should prevent some inevitably-cancelled UI
thread tasks from being posted on a dead process's behalf.

Bug: 1000327
Change-Id: Ib2701e2d99a922c991d57745ad63a6d9b8c294ad
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1810036Reviewed-by: John Abd-El-Malek <jam@chromium.org>
Commit-Queue: Ken Rockot <rockot@google.com>
Cr-Commit-Position: refs/heads/master@{#697536}