DescriptionAdd an experiment to redirect DOMStorageTaskRunner to TaskScheduler.
Using a bug as an opportunity to experiment with TaskScheduler (as well
as confirm the issue is what we think it is: a priority inversion where
a sync IPC from the renderer is blocking on a response from the
BlockingPool).
Only logical difference is that TaskScheduler doesn't have a notion of
named sequences and as such every DOMStorageContextWrapper will have its
own primary and commit sequence instead of sharing one with other
DOMStorageContextWrapper. If anything that's a good thing as it allows
multiple profiles to do their DOMStorage ops in parallel? Don't think
this can conflict..?
BUG=665588, 622400
Committed: https://crrev.com/72b1a059ae930ebe1dda804e607f251d3ae94a70
Cr-Commit-Position: refs/heads/master@{#439575}
Patch Set 1 #Patch Set 2 : typo in comment #Patch Set 3 : rebase on r438937 #Patch Set 4 : update dependencies #Patch Set 5 : rebase on r439147 #Patch Set 6 : update dependencies #Patch Set 7 : fix compile #Depends on Patchset: Messages
Total messages: 43 (33 generated)
|