Fix lyric extenders to end properly when there are notes but no more lyrics.

Fix lyric extenders to end properly when there are notes but no more lyrics.

As I understand it, the current extender code only ends extenders when the next
lyric is encountered. This can lead to issues when there are additional notes,
but no more lyrics (I encountered this case when I needed to add a few lyrics
where rhythms diverged in a divisi part). This patch adds a check to see if a
melisma is ending, completetizing the extender if appropriate.

This has solved the issue I was encountering, and the regression-test turned up
no regressions. A test-case with before-and-after output is available at the
following URL:
http://temp.mvpsoft.com/ly/lyric_extender/

This also seems related to Issue #331, which mentions a similar problem. When I
tried the test-case included there, however, it produced correct output with and
without this patch.

[joeneeman, 2009-03-16 22:02:55]

Here are some code-style nitpicks. Also, please add a regression test that
demonstrates the change (a minimal example, smaller than the one you placed on
your website).

http://codereview.appspot.com/28052/diff/1/2
File lily/extender-engraver.cc (right):

http://codereview.appspot.com/28052/diff/1/2#newcode101
Line 101: if (! melisma_busy(voice))
Code style is:
if (!melisma_busy (voice))

http://codereview.appspot.com/28052/diff/1/2#newcode102
Line 102: {
Indent the brace by 2 spaces.

[csnyder, 2009-03-16 23:32:20]

On 2009/03/16 22:02:55, joeneeman wrote:
> Here are some code-style nitpicks.

Fixed (though the rest of the file is still quite messy).

> Also, please add a regression test that demonstrates the change (a > minimal
example, smaller than the one you placed on your website).

Done (though the difference in the extender line length wasn't enough to cross
the distance threshold in the regression-test).

[joeneeman, 2009-03-17 00:05:17]

On 2009/03/16 23:32:20, csnyder wrote:
> On 2009/03/16 22:02:55, joeneeman wrote:
> > Here are some code-style nitpicks.
> 
> Fixed (though the rest of the file is still quite messy).

Sure, but there's no point making it worse :)

> > Also, please add a regression test that demonstrates the change (a > minimal
> example, smaller than the one you placed on your website).
> 
> Done (though the difference in the extender line length wasn't enough to cross
> the distance threshold in the regression-test).

The regression testing uses ratios between areas, which doesn't seem to work
well for very thin and elongated symbols. I'll see if something can be done.

If no one else comments, I'll push this in a day or so. For future reference,
btw, please include lilypond-devel@gnu.org in the CC list of issues so that
these discussions are more visible.

Thanks for the patch,
Joe