Are there conceivably cases where ignoring EIO as a non-failure will /not/ do the right thing here, and instead cause a regression?
Accepting into -proposed, but I would appreciate seeing some discussion of how we know this won't cause regressions for users not affected by the current problem before it goes to -updates.
Are there conceivably cases where ignoring EIO as a non-failure will /not/ do the right thing here, and instead cause a regression?
Accepting into -proposed, but I would appreciate seeing some discussion of how we know this won't cause regressions for users not affected by the current problem before it goes to -updates.