Re: Planning for 2.22.2

[Jean Abou Samra]

Le 10/01/2022 à 20:35, Jonas Hahnfeld via Discussions on LilyPond 
development a écrit :
> Hi all,
>
> there's a number of fixes stacking up in stable/2.22 (at least after I
> processed the patches that I marked for backporting in the past months)
> and I'd like to tentatively plan for the next bug fix release 2.22.2
> some time in February. We could collect some bonus points by going for
> the 22nd, but we'll see if and how that works out 😉
>
> So please let me know if you have anything that you think would be good
> to include and that I haven't backported yet. I hope wrote a comment on
> all MRs that I cherry-picked commits from, and changed the milestones
> of the corresponding issues, if there is one...
>
> I would have liked to include some more balloon fixes, but already
> https://gitlab.com/lilypond/lilypond/-/merge_requests/1071 doesn't
> produce the expected results (no footnote on the MultiMeasureRestNumber
> and \after is not available in 2.22 at all),


Right, I didn't realize that. I think it's a case
fixed by

commit 4f474c99d1f7afdec28a159bf1ccc99d619c7dc7
Author: Jean Abou Samra <jean@abou-samra.fr>
Date:   Mon Sep 20 21:11:37 2021 +0200

    Let get_bound default to host's bound for sticky spanners

    This solves all problems related to the host's right bound being unset
    at the point where the Spanner_tracking_engraver tries to pass it 
on to
    the attached sticky grob.  Since bounds can be set as late as in
    finalize, the Spanner_tracking_engraver is repurposed to only announce
    end, and the bounds are now handled by defaulting to the hosts' 
when they
    are retrieved.

    Closes #6158, #6159

but that one touches a core part of the code, so
I wouldn't advise backporting it. Also, !1071
is likely not to fix an actual problem if you
don't backport

commit ac703cc5b7d8c54650ac36677243799bf2772169
Author: Jean Abou Samra <jean@abou-samra.fr>
Date:   Tue Sep 21 00:39:57 2021 +0200

    Set X parent of sticky spanner to host

    In the next commit, the bounds of sticky spanners will no longer be 
set
    explicitly, so they will no longer get an X parent automatically.  
Give
    them their host as parent.  This is actually more likely to match
    expectations than the previous behavior of making the left bound X
    parent.

which is just the previous commit. All in all
a long story with lots of fixes interacting
in complex ways if you leave out some of them, so
I'd ultimately recommend not backporting any of that.


> and the fix from
> https://gitlab.com/lilypond/lilypond/-/merge_requests/1077 doesn't seem
> to work either.


It will fix footnotes, but not balloons because you'd
most likely also need

commit 10e8952de0494a65323e9d9b2ecef129c7f7f8ca
Author: Jean Abou Samra <jean@abou-samra.fr>
Date:   Sat May 22 11:36:47 2021 +0200

    Fix balloons on grobs with empty pure Y extent

    This is done by introducing a new function, 
robust_relative_pure_y_extent,
    which is the pure equivalent of robust_relative_extent.  When the 
extent
    is empty, it returns a single-point interval rather than the empty 
interval.
    This is accurate for balloons, because they use robust_relative_extent
    in the unpure stencil.

    Closes #5306.

Wait, actually I see it in the stable/2.22 branch,
and if I backport d22f1bb88e manually, it seems
to work. Not sure how you've tested it?

At any rate, even if the spacing differences that
d22f1bb88e causes for glissandi where the Y endpoints have been
overridden explicitly are good-natured, they're a bit
more than we'd want to put in a point release I think.
Again, I would not recommend backporting this one.
No strong opinion here though.

Thanks for working on the release!

Jean