Inconsistent placing of TrillSpanners when not using \stopTrillSpan

[James Lowe]

Hello,

While updating the NR with the reg tests that have not yet been 
documented I came across an anomaly with trill spanners.

Specifically the reg test says

> Consecutive trill spans will work without explicit \stopTrillSpan commands, 
> since successive trill spanners will automatically become the right bound of 
> the previous trill.

Here is my simple ly file showing you 3 different scores, the only 
difference being the \relative c.

%%%

\version "2.13.24"

\score {
  \relative c''' {
    c1\startTrillSpan
    c1
    a1\startTrillSpan
    c1\stopTrillSpan
  }
}

\score {
  \relative c'' {
    c1\startTrillSpan
    c1
    a1\startTrillSpan
    c1\stopTrillSpan
  }
}

\score {
  \relative c' {
    c1\startTrillSpan
    c1
    a1\startTrillSpan
    c1\stopTrillSpan
  }
}

%%%

I have attached a jpg for those that don't want to run the lily example 
directly.

What I don't understand is why the middle example shows that the second 
trill allows for the 'overlap' but the first and third don't and the 
trills effectively run into each other.

I'm not looking for a solution for documentation - I can deal with that, 
but wondered if this was a 'bug' or 'expected' (if you see what I mean?).

James

Capture.JPG
Description: JPEG image