Re: Lyrics break estimation of vertical spacing

[Boris Shingarov]

On Wed, 24 Mar 2010 14:19:21 -0700, Joe Neeman  wrote:
> Right, but in align-interface.cc:118, we throw away that unite()d extent
 > in favour of a begin_of_line_extent and a rest_of_line_extent.
 
But that, IIUC, is only happening one level (of the grob tree) above,
when it's too late, no?  That is, inside Grob::pure_relative_y_coordinate()
(grob.cc:456), which follows the Guile call, the returned value from which
is already the unify()'ed value, and then we translate *that* value (already
too big).
I suspect that *may* be where the bug is, but I am not sure.
The end result of severe overestimation of height when low-hanging
rest-of-line elements are present (such as lyrics or low notes), has been
there for a very long time.
To me, the whole mechanism of beggining-vs-rest-of-line, is exactly
about it, so that means that the whole mechanism is broken (or is there
an example where it is working?)  But I find it hard to believe that we
have been content with living with it for so long.