Pedro J. Estébanez

1. Any comments on my suggestion about using `reference->unreference()` instead of the cleanup ref approach? 2. OK. - (We'd actually have to judge on each condition separately. Maybe some of...

Such a change, despite indeed making performance better, makes the code much less maintenable due to the hardcoded size and indices. The right way of addressing this would be the...

> > The right way of addressing this would be the same in, for instance, > > I know nothing about it but should it actually be cleared and constantly...

I see you're introducing a new convention for template arguments (`t_` prefix). Not a big deal since we don't have uniformity in that regard across the codebase. However, it may...

I've found something a bit concerning... One of my PRs was tested against 4.3, where it worked. On _master_ it broke. The reason was that on _master_ I had to...

For what it's worth, and not necessarily in the scope of this PR, but on this topic, it'd be great that there was some automatic layout logic so the wrapping...

I think this makes a lot of sense. When you are targeting very different devices, it's a very natural thing to do. I remember when back in time you'd bundle...

Makes sense. I have a little concern though. Is the former value for "all devices" stored somewhere? In that case, it'd be appropriate to leave `-1` for that.

Why don't just leave `-1` for "all devices" and use, say, `-2` for emulation?

So, being that the case, your proposed workaround seems to be the most sensible thing to do.