Why? The directory is always just "native" (one level)

I agree with Igor.  If -p is passed no error will be produced if the directory alreay exists.

I didn't remember that -p had that side-effect.  OK.

You're both wrong, you want $(INS.dir).

Sorry, I only just noticed that RB swallowed my response to this.  To clarify, the reason you want to be using $(INS.dir) instead, is uniformity with every other directory we create (it uses install under the hood, and is also safe for the same reason as -p).

You sure about that?  I see a bunch of -@mkdir -p in the gate too.

I'd have guessed INS.dir would be for things that we're actually

installing (in the proto area) which is not what's going on here.

If we're already sloppy about it, do either.  I just remember having being asked to do it that way before :)

Added that to the bug.  (the "native" directory may already exist, i.e. in an incremental build)