Re: Issue 1771 in lilypond: Nested reverts fail for properties which aren't set as defaults in define-grobs.scm

[lilypond]

Updates:
        Status: Duplicate
        Labels: -Priority-Critical Priority-Medium
        Mergedinto: 1063

Comment #3 on issue 1771 by address@hidden: Nested reverts fail for  
properties which aren't set as defaults in define-grobs.scm
http://code.google.com/p/lilypond/issues/detail?id=1771

This one is likely Invalid.  1063 is not possible to fix without breaking  
other aspects of nested properties -- they are broken by design.  You can't  
make an alist one property and transparently have the full  
context-hierarchic override/revert functionality for its sublevels  
independently.

The notation manual (node "Modifying alists") is not particularly  
informative regarding the fine details.  It just mentions that using nested  
properties will not work for properties which are themselves alists.  But  
as far as I can see, this is exactly what the implementation of nested  
lists is now.  So it is my understanding that it will be easy for any  
implementation relying on the current design of the data structure and the  
transparently accessible alists to write a regtest that fails.  More  
obfuscated code, like the one in the patch leading to this regression, will  
just make it harder to predict the behavior and create a matching failing  
regtest.

Since the report is not technically invalid, I merge it into 1063.  The  
error symptoms are, of course, different, but both, to my understanding,  
can't be fixed simultaneously, at least not without breaking functionality  
at yet another place.  So they are both a result of the same design error,  
and need a common strategy to tackle.

Feel free to prove me wrong.