Creates a LeftBarStub that stops lyrics from bumping into the SystemStartBar.

Creates a LeftBarStub that stops lyrics from bumping into the SystemStartBar.

[pkx166h, 2011-10-06 19:41:26]

passes make but I get a programming error show up on make check (as well as half
a dozen reg test diffs).

See attached

http://code.google.com/p/lilypond/issues/detail?id=1956#c1

[MikeSol, 2011-10-07 09:47:03]

On 2011/10/06 19:41:26, J_lowe wrote:
> passes make but I get a programming error show up on make check (as well as
half
> a dozen reg test diffs).
> 
> See attached
> 
> http://code.google.com/p/lilypond/issues/detail?id=1956#c1

All these are fixed.  Now, the only diffs should be lyrics that used to be flush
with the system start being pushed to the right by 1.0 units.

[pkx166h, 2011-10-08 04:59:31]

Passes make and make check shows 2 reg test diffs that Mike seems to think are
expected

attached at:

http://code.google.com/p/lilypond/issues/detail?id=1956#c3

[Neil Puttock, 2011-10-08 16:41:52]

http://codereview.appspot.com/5201043/diff/3001/scm/output-lib.scm
File scm/output-lib.scm (right):

http://codereview.appspot.com/5201043/diff/3001/scm/output-lib.scm#newcode355
scm/output-lib.scm:355: (if (eq? 0 (ly:item-break-dir grob))
Doesn't this potentially make the function unpure?

If I compare Item::break_visible ()/Item::pure_is_visible (), the main
difference is the latter doesn't call break_status_dir ().

[MikeSol, 2011-10-10 10:29:26]

This new patch set eliminates all regtest diffs via a mechanism that allows
extremal lyrics to be flush with the left edge.
It kinda sorta bends LilyPond into doing something it is not designed to do ::
returning pureY-extents for items that are dependent on the start and end
positions (normally, only spanner Y-extents change based on start and end
positions).  The logic is that a LeftBarStub should have a Y-extent iff it is at
the beginning of a staff and if said staff will not bump into a SystemStartBar. 
Knowing if it will bump into a SystemStartBar can only be done if we know which
staves are alive and which staves are hara-kiri'd, which'll change based on the
start / end in question.

This seems like the sorta thing that'd slow down lilypond, so I'd appreciate it
if people tested it on pieces with lots-o-lyrics!

Cheers,
MS

[MikeSol, 2011-10-10 11:13:00]

I got a little ahead of myself...the patch, while it works in the regtest cases,
does not work the way I think it did because all Items' pure heights are
calculated from 0 to INT_MAX in Separation_item::boxes.

Certain NonMusicalPaperColumns' horizontal skylines will change depending on if
they begin a line or not.  It seems, though, that Separation_item does a
one-size-fits-all calculation for the minimum distance between
NonMusicalPaperColumns and PaperColumns.  So, my questions are: do
NonMusicalPaperColumns have different horizontal skylines for different break
visibilities?  If so, where are they stashed?  If not, how does the horizontal
spacing engine know how to deal with changing break visibilities for different
start and end columns?  I'm thinking that Line_shape may have something to do
with it, but I'm not sure.  Help would be appreciated!

Cheers,
MS