The commit comment should have a [section.label] prefix.

Other than that, I'm not sure what this footnote delivers us in the first place: Reading uninitialized data is undefined behavior, and attempting to list the shades and variations of undefined behavior leads nowhere at best and is seriously misleading otherwise. So, let's remove that footnote altogether.

See [dcl.init] p12 for the detailed treatise. I'd actually be in favor of moving most of this paragraph into [expr.prop] as a new subclause "Indeterminate values". Except that an indeterminate value is not a compile-time property (like the other properties mentioned there), but a runtime property. @zygoloid, what do you think?

I think this change makes matters a little worse, since it makes this note sound even more like it's describing what will actually happen. I think this note is problematic for other reasons, as (by omission) it suggests that the same might not be true for other types. (Eg, reading an uninitialized int is undefined even if there's a one-to-one correspondence between object representations and value representations.)

I'm in favor of deleting this footnote. Moving the description of uninitialized values into its own subclause seems very reasonable to me, but it's not a property of an expression, so I don't think [expr.prop] is an ideal home. Maybe [intro.object], perhaps just after [basic.life], would be better? I think we should move all but the first sentence of [dcl.init]p12 to the new subclause, and add a paragraph break after the first note.

The note is now gone. 👋