jamescook@, I am pretty sure this will happen because the update state exists only in memory, so if Chrome is restarted it forgets that there was an update. How fixable is this?

Note: I also need to test if logging out triggers a reboot in this case.

I guess you'll need some way to query if there's a pending Flash update when chrome starts? I'm not sure how hard that is.

If there's an easy way to do that then the SystemTrayClient / SystemTrayController could push that information down into ash.

Thanks I will investigate. In terms of the priority of this, how many users do we think will actually type chrome://restart into the browser? I suspect it is not a common thing to do.

It seems pretty rare.

If the only way to fix this is to do disk IO during startup / login I might just decide not to do it.

I wonder if update_engine can periodically tell Chrome that there's still an update pending. I don't know if it just does it once.