[Dwarf-discuss] RE: 2.12.1, Non-defining Declarations [Dwarf-discuss Digest, Vol 9, Issue 3]

[Chris Quenelle]

> Cary Coutant's note deserves calling out in the 
> specification, I think: shouldn't it be a meta-rule
> that every attribute has a value, and no attribute's
> "mere presence" is significant?
> 

I'd like to see a significant discussion (in rationale)
of this issue, because without such a discussion
the specification is a little confusing.  It's obviously
redundant, so why did we make it that way.
John's summary seems like a good starting point for the rationale:

------

As specified, producers have a choice of three reasonable
methods:

1.  Always have a DW_AT_declaration attribute and
     set the flag true or false;

2.  Have two different abbreviations, one of which
     has the attribute; uses of the abbreviation
     with the attribute will always have a true value
     (redundantly).

3.  Have two different abbreviations, one of which
     has the attribute; uses of the abbreviation
     with the attribute may have true or false flags.