Code review comment for lp:~mir-team/mir/improve-attribute-protocol-and-normalize-getters

Yay! Surface attributes populated on construction! I'm broadly in favour of this :)

Your API docs don't match the behaviour; they say you'll return -1 for invalid surfaces.

While we're at it, I thought that, waaaaay back in London, we were going to abort() when passed an invalid surface? I'm very much against silently pretending that everything's OK. Likewise, I cringe at all the

try { <something> }
catch(...)
{
    //la di da! The client doesn't need to know about this...
}

in there (even though much of it is pre-existing).