On Thu, Jun 24, 2021 at 03:40:21PM +0100, Mark Rutland wrote:

> regular unwinds (e.g. so that we can have a backtrace idicate when a
> step is not reliable, like x86 does with '?'), and to do that we need to
> be a little more accurate.

There was the idea that was discussed a bit when I was more actively
working on this of just refactoring our unwinder infrastructure to be a
lot more like the x86 and (IIRC) S/390 in form.  Part of the thing there
was that it'd mean that even where we're not able to actually share code
we'd have more of a common baseline for how things work and what works.
It'd make review, especially cross architecture review, of what's going
on a bit easier too - see some of the concerns Josh had about the
differences here for example.  It'd be a relatively big bit of
refactoring though.