Re: 2.15.41 Regtests

[David Kastrup]

"Phil Holmes" <address@hidden> writes:

> I've been doing a bit of work with my pixel comparator, and now have a
> mechanism to generate a web page from commented results.  Running the
> checker on 15.41 versus 15.40 gives about 100 changed files, most of
> which are subtle changes, some of which come from a changed forte
> mark.  I've got a web page with what I believe are significant
> changes.  Its address is:
>
> http://www.philholmes.net/lilypond/regtestresults/
>
> Please look at it and see if these are intended changes.

Responsible likely

commit 16a328617e867892e8d5608f95bd6127196e03a2
Author: Keith OHara <address@hidden>
Date:   Sun Jun 17 21:45:26 2012 -0700

    Consistent bounds for slurs; issue 427 and 379

commit 619997f335c96cb3720f486650fd284906c90b83
Author: Keith OHara <address@hidden>
Date:   Sun Jun 17 21:43:54 2012 -0700

    phrasing-slur: avoid key-sigs, clefs; issue 49

commit ad0e41c6cd25db2643d1a3e25242d284d6f65e69
Author: Mike Solomon <address@hidden>
Date:   Fri Jun 15 08:37:47 2012 +0200

    Forbids slur and tie extrema points from getting to close.


for the slurs.  There is no commit message that would explain the
shortened stems like in

input/regression/collision-head-solfa-fa.ly 

but it is easy to verify that the following commit is responsible:

commit 30f2bb655e6f240a795af9f52c1476849b378ead
Author: Mike Solomon <address@hidden>
Date:   Fri Jun 15 08:27:17 2012 +0200

    Gives stencilless unbeamed stems an empty Y-extent.


It is obvious that LilyPond's idea of "stencilless" is different from
Mike's since I verified that

  /*
    If there is a beam but no stem, slope calculations depend on this
    routine to return where the stem end /would/ be.
  */
  if (!beam && !unsmob_stencil (me->get_property ("stencil")))
    return Interval ();

in lily/stem.cc is responsible for the regression.

-- 
David Kastrup