On Sun, Oct 7, 2018 at 10:56 PM David Gibson <david@xxxxxxxxxxxxxxxxxxxxx> wrote: > > On Sun, Oct 07, 2018 at 04:46:13PM +0200, Lubomir Rintel wrote: > > This would catch the cases, where the output DTS still parses well and > > produces a tree that we expect, yet it contains some weirdness. > > I'm really not sure what sort of thing you're trying to catch with > this. The problem is that exactly how we format the dts is > necessarily a heuristic - if we adjust or improve how we guess the > formatting for dts output the output *will* change, but it won't be > incorrect. Yes, but we've obviously had regressions here because the dts varies now based on the input source. It's not just checking for correctness, but unintentional changes. If what we intend for dts output changes, then the test data should too. Rob
