David Kastrup [Sat, 4 Jun 2016 12:11:31 +0000 (14:11 +0200)]
Issue 4997/2: Use Preinit class in Scheme_engraver
This avoids calls of derived_mark on uninitialized data. Fold
Scheme_engraver::init_from_scheme (SCM) into converting constructor
since it has become identical to it.
David Kastrup [Sat, 4 Jun 2016 12:09:08 +0000 (14:09 +0200)]
Issue 4997/1: Add Preinit class for early initialization
This is a tricky thing: once a base class calls smobify_self () in
its constructor, further allocations during construction of base
class and derived classes might lead to mark_smob calls on the
object under construction. When those call a virtual function like
derived_mark, the virtual function corresponding to the
incompletely initialized object is likely to be called.
This situation is acerbated in Guile 2.0 where the mark passes may occur
with only loose synchronization to the allocations.
The order of initialization of an object consists in calling the
constructors of virtual base classes, then of non-virtual base
classes, then initializing all data members.
As a result, the derived constructor comes too late for
initialization. That's where the Preinit template class comes in.
Derive from it _before_ deriving from the smobifying base class
providing derived_mark, and it will call its Base class' pre_init
function (which must not rely on the instantiation being complete).
The pre_init function should then put the derived class into a state
_before_ its constructor or a constructor of its data members is being
called where it's safe to call its derived_mark function. Note that in
particular that most STL data structures are in undefined state before
their constructor is being called, so the pre_init function has to
establish a state where derived_mark will not attempt to interpret data
members of types like std::vector which may not have been constructed
yet.
Masamichi Hosoda [Fri, 28 Oct 2016 13:53:25 +0000 (22:53 +0900)]
Issue 4991: Add installing optional URW++ fonts
LilyPond default fonts (TeX Gyre) do not have Greek and Cyrillic glyphs.
Newest URW 35 fonts (June 2016) have them.
http://git.ghostscript.com/?p=urw-core35-fonts.git;a=commit;h=79bcdfb34fbce12b592cce389fa7a19da6b5b018
This commit realizes the followings.
. configure script finds the following 12 OTF files.
. If they are found, `make install` installs them.
(Both TeX Gyre and URW fons are installed
under the lilypond fonts directory.)
. Even if they are not found, configure script does not raise error.
(Only TeX Gyre fonts are installed.)
Antonio Ospite [Wed, 9 Nov 2016 17:22:25 +0000 (18:22 +0100)]
Fix ending the dynamic extent in Text_interface::interpret_markup
When using cyclic references under guile-2.0 lilypond crashes with
a segmentation fault:
$ LANG=C out/bin/lilypond input/regression/markup-cyclic-reference.ly
GNU LilyPond 2.19.51
Processing `input/regression/markup-cyclic-reference.ly'
Parsing...
Finding the ideal number of pages...
Fitting music on 1 page...
Drawing systems...
Layout output to `/tmp/lilypond-wkUlSF'...
Converting to `markup-cyclic-reference.pdf'...
Deleting `/tmp/lilypond-wkUlSF'...
Segmentation fault
This happens because in Text_interface::interpret_markup the dynamic
extent is not ended properly in the exit path for the case when the
markup depth exceeds the max_depth.
Masamichi Hosoda [Fri, 28 Oct 2016 14:11:29 +0000 (23:11 +0900)]
Issue 4992/2: Add `never-embed-font-list` and `/NeverEmbed`
When the option `-dgs-never-embed-fonts` is enabled,
`/NeverEmbed` distiller parameter from `never-embed-font-list` list
is added to the outputting PostScript file.
When Ghostscript converts from the PostScript file to PDF file,
it never embed the fonts which are in the `never-embed-font-list`.
Thomas Morley [Tue, 25 Oct 2016 20:29:20 +0000 (22:29 +0200)]
Issue 4988 straight-flags partly wrong calculated
The angles for straight-flags at down- and up-pointing stems are now
accurately respected.
This will cause some slight changes in the reg-tests.
Also fixing flat-flags, which relied on the old, buggy calculation.
Also doing the same fixes for markup-command note-by-number.
Also fixing an oversight in regtest metronome-mark-formatter.ly,
where the creation of the metronome-markup depends on note-by-number.
Paul Morris [Mon, 29 Aug 2016 04:27:19 +0000 (00:27 -0400)]
Issue 4962: Web: home page: add example image, reduce news to headlines
The example image is the excerpt from Bach BWV 861 from the essay.
Anchor links are used to link to specific examples on the examples
page. The news entries on the home page are reduced to just a list
of headlines that link to specific entries on the 'Old News' page
(under 'Community'). And that page is renamed to just 'News'.
Mark Knoop [Thu, 8 Sep 2016 17:56:16 +0000 (18:56 +0100)]
Keep a staff alive with multiple layers
This allows the `VerticalAxisGroup.remove-layer'
property to accept a list of values. The layer
will stay alive with any other member of the
Keep_alive_together_engrave group with a
remove-layer value in that list.
The principal reason for this patch was to
allow the use of MarkLine contexts in a Frenched
score, where the context should stay alive with
any single staff in a StaffGroup.
This implementation should also allow additional
flexibility with ossia and divisi staves.
David Kastrup [Fri, 2 Sep 2016 21:11:53 +0000 (23:11 +0200)]
Issue 4965: Create and use Grob::parent_relative
This function checks for the existence of a Grob parent before
calculating a coordinate relative to it. This should hopefully
clean up the most relevant problems caused by issue 4814
and the original GCC 6 optimization causing it.
projects / lilypond.git / log