@Oli any idea why the miri-rustc-tests binary passes
nope, have you tried git blaming the line?
oh. right. a few PRs before that we were manually registering the optimization passes
which caused us to diverge from the rustc passes
so this change was to not change behavior
seems odd though, to make level 3 the default
I think level 0 should be default, and it actually is in my latest miri branch
and in making that consistent I realized this oddity in the rustc test runner
I'd prefer to remove it, make things more consistent?