Re: Uses derived_mark to avoid segfault (issue 5729051)

[mtsolo]

http://codereview.appspot.com/5729051/diff/1001/lily/span-bar-stub-engraver.cc
File lily/span-bar-stub-engraver.cc (right):

http://codereview.appspot.com/5729051/diff/1001/lily/span-bar-stub-engraver.cc#newcode71
lily/span-bar-stub-engraver.cc:71: axis_groups_ = scm_cons (scm_cons
(i.grob ()->self_scm (), i.context ()->self_scm ()), axis_groups_);
On 2012/03/04 07:26:30, dak wrote:
> axis_groups_ never gets cleared out again: this keeps grobs and
contexts alive
> at least until the engraver is collected.

This is the point.  I was wrong to suggest before that it be cleared
out.
Unless the score is insane, a context like PianoStaff will only ever
house 10ish contexts and 10 vertical axis groups max.  So the size will
stay tiny.

I could clear this out, but the problem is that LilyPond doesn't provide
a mechanism to signal when VerticalAxisGroups are no longer being used.
If announce_end_spanner were called on them, I could acknowledge that in
the engraver.  A separate tracker issue could be opened for that.

http://codereview.appspot.com/5729051/diff/1001/lily/span-bar-stub-engraver.cc#newcode81
lily/span-bar-stub-engraver.cc:81: if (!vertical_alignment) // we are at
the beginning of a score, so no need for stubs
On 2012/03/04 07:26:30, dak wrote:
> If "we are at the beginning of the score" is a valid state here,
scm_caar
> (axis_groups_) is likely to throw an exception.

This is not possible because process_acknowledged is called after all
other process calls, so the list will be populated.  However, I agree
that an extra check here couldn't hurt.

http://codereview.appspot.com/5729051/diff/1001/lily/span-bar-stub-engraver.cc#newcode127
lily/span-bar-stub-engraver.cc:127: for (vsize j = 0; j <
affected_contexts.size (); j++)
On 2012/03/04 07:26:30, dak wrote:
> While you are at it: how about throwing out all of the reverses and
instead
> letting the loop run backwards?

This is vestigial code from when I was doing sorting on the span bar
vector.  It is no longer necessary, so removed.

http://codereview.appspot.com/5729051/