Re: tie direction !?

[David Kastrup]

Torsten Hämmerle <address@hidden> writes:

> Thomas Morley-2 wrote
>> I already thought creating a function for setting tie-direction, _iff_
>> the notes are on the middle line, probably taking stem-direction into
>> account. It would give me more flexibility, though ofcourse the
>> considerations mentioned above are still valid.
>
> Actually, the C++ coding does take the stem directions of the tied notes
> into account, but in our cases, they point in opposite direction and so any
> tie direction will be equally good or bad.
>
>
>
> Thomas Morley-2 wrote
>> Obviously it has something to do with the notes being on the
>> middle-line, so I tried the some inconsistent behaviour with notes on
>> other lines in order to get more info/insight. Without success so far
>
> Yes, that's why I claimed this strange effect only happens for middle-line
> ties.
>
> The up/down decision seems to be fragile as hell.
>
> As it's a centre-line (i.e. Y-offset = #0) problem, this reminds me of
> numeric rounding errors (inaccuracies): sometimes, the result is < 0,
> sometimes it's > 0, but always very close to 0.

Can be a memory-order thing.  When two things compare equal, their final
order of decisions can depend just on which choice happened to get a
location lower in memory.  In that case, the results need not even be
deterministic given identical scores on the same platform with the same
program.

-- 
David Kastrup