Re: Possible regression with box-markup/rounded-box-markup

[Thomas Morley]

2018-03-28 21:09 GMT+02:00 Torsten Hämmerle <address@hidden>:
> In any case, the additional space around the \rounded-box is proportional to
> the corner-radius.
> If you set corner-radius to 0, there's no gap anymore. As the corner-radius
> increases, so does the gap.
>
> I think it's the rounded-box-stencil's dimensions:
> when explicitly setting them to xext/yext, everything works as expected (w/o
> the extra gaps).
>
> I suspect ly:round-filled-box, as rounded-box-stencil itself hasn't changed.
>
> All the best,
> Torsten



Hi Torsten,

thanks for your findings.
I meanwhile identified the first bad commit:

commit 28f3294954eff1f263d3b2e3de1c520f4d2fbdfc
Author: Mike Solomon <address@hidden>
Date:   Mon Aug 27 23:47:04 2012 +0200

    Improvements in vertical skyline approximations (issue 2148).

    The file stencil-integral.cc provides a suite of functions that
    traverse a stencil and do linear approximations of its components.
    These are then turned into boxes that are passed to the Skyline
    constructor. This approximation is used for several vertical skylines
    including those of VerticalAxisGroup and System. As a result of these
    more accurate approximations, vertical spacing is more snug between
    grobs.

    Additionally, in axis-group-interface.cc, skylines of grobs are no
    longer compared to a monolithic axis-group skyline but rather all
    of the component skylines of the axis-group, allowing grobs to
    be fit under other ones if there is space instead of always shifted over.

    Two new python scripts allow to visualize the position of skylines.

    All other changes provide functions that allow for better debugging
    of Skylines, better approximations of grobs via skylines, and changes
    to the measurement of distance between grobs via the new Skyline API.

    This results in a significant time increase in score compilation for
    objects with complex skylines such as all text grobs.  For orchestral
    scores, the increase is not as steep.

https://sourceforge.net/p/testlilyissues/issues/2148/
https://codereview.appspot.com/5626052/

Unfortenately this is Mike's major skylines-patch.

With a quick glance over it, I found some code for
make_round_filled_box_boxes
Probably  a point to look at, but it's mostly work in C++, so I'm off :(

I had to revive old LilyDev3 to compile old (2.17.) versions, i.e. with
address@hidden ~$ gcc --version
gcc (Debian 4.7.2-5) 4.7.2
(in case someone wants to recheck)


Thanks,
  Harm