Re: vertical movement without anchors

[David Kastrup]

"Phil Holmes" <address@hidden> writes:

> ----- Original Message ----- 
> From: "David Kastrup" <address@hidden>
> To: "Werner LEMBERG" <address@hidden>
> Cc: <address@hidden>
> Sent: Monday, July 25, 2016 2:07 PM
> Subject: Re: vertical movement without anchors
>
>
>> Werner LEMBERG <address@hidden> writes:
>>
>>>>> When has this changed?  Or maybe there are still situations where an
>>>>> anchor is needed, thus the example has to be improved?
>>>>
>>>> The respective diff from 2.16 to 2.18 in scm/define-grobs.scm reads:
>>>> [...]
>>>>
>>>> However, reverting all that does not appear to make a change (unless
>>>> I am doing something wrong here).  So I'm not sure what to attribute
>>>> the change to (the logs for lily/text-engraver.cc or
>>>> lily/script-engraver.cc do not show anything suspicious to me
>>>> either).
>>>
>>> Thanks for testing.
>>>
>>>> How important is it to figure out the responsible change?
>>>
>>> Well, it's not important (at least not for me), but `anchors' was a
>>> central concept in vertical positioning of grobs, and lo and behold,
>>> it is no longer necessary, so we can correct the documentation.
>>
>> Could have something to do with issue 3330, committed as part of
>> 2.17.19.  No idea just _what_, though (but it streamlines and amends the
>> spacing engine considerably, so it would be a likely candidate).  Do we
>> have the docs for 2.17.18 and 2.17.19 for comparison?  In other words,
>> I'm a bit lazy.
>>
>> -- 
>> David Kastrup
>
> Happened between 2.17.26 and 2.17.28

Ok...

I suspect

commit 2c0f56cc2c2b7e6d3a5f7792c006e55675c9e0d7
Author: Keith OHara <address@hidden>
Date:   Sun Dec 23 13:13:44 2012 -0800

    Measure 'staff-padding' to reference points; issue 3026


Reverting its C++ part (the merge conflict in the Scheme part in
scm/define-grobs.scm would be rather messy to clean up) in
lily/side-position-interface.cc (and a followup commit for issue 3626
removing a then-unused variable) results in output corresponding to the
old situation.

And the issue description very much matches the symptoms.

The salient difference in the NR example is that the \null addition
changes the bounding box (to include (0,0), the reference point) but not
the reference point, and Keith's commit causes the padding to be
measured to the (unchanged) reference point rather than the (changed)
bounding box.

So this does not really have anything to do with "anchoring markups" as
far as I can see.  Unless you consider a markup "anchored" when its
bounding box and/or skyline includes the reference point.

-- 
David Kastrup