Minutes, Dwarf 2 Meeting - December 7, 1999
It was generally felt that this proposal needed elaboration.
There was extensive discussion of namespaces, their complexity, and
how they might be described. Questions were raised about whether a
namespace was a scope, and whether it needed to be described at all.
A compiler resolves references to symbols using namespace and other
information. These symbols are fully resolved in Dwarf.
Describing namespaces in Dwarf would allow a debugger to resolve
references to symbols which were not referenced in the compiled code.
This raised a concern that the debugger would have to resolve names
identically to the compiler, which may be problematic.
Jason Merrill will send an example of describing a namespace using
this proposal. Rajiv & Bevin also indicated that they would send
comments and/or alternative descriptions.
991129.1 Pubnames interpretation
The Dwarf spec uses Compilation Unit Entry in several places, notably
6.1.1 and 6.1.2, in ways which are open to different interpretations.
After discussion, it was decided that this is an editorial issue and
does not require any change to the Dwarf definition. Wherever there
is a reference to the offset from a Compilation Unit Entry the intent
is that this be the offset from start of the Compilation Unit Header.
The wording should be clarified, perhaps to define a Compilation Unit
as the Header followed by the DIEs for the CU.
991026.5 Register overloading
Withdrawn. Unclear who raised this issue, or what was intended.
991108.10 Array bounds
Dwarf uses the Subrange type for two purposes: to describe range
values (such as in Pascal) and to describe array ranges. The
specification says that if either bound is missing, the value is
a language-specific constant. This appears to be reasonable for
the use as an integer range or arrays where the size is known, but
not for arrays where the dimensions are unknown. The comment in
5.10 (which is explanatory and not part of the specification)
suggests that no default is provided for C/C++ for the upper bound.
This makes describing arrays where the upper array bound is not
known unclear, since no constant upper bound would be appropriate.
It was decided that rewording the comment in this section, as well
as possibly adding a comment in 5.4 describing Array Types, would
resolve this issue. Jason Merrill will propose wording.
There had been some discussion on the reflector about extensions
to support assembly language(s). No written proposal has been
received on this issue. It has been suggested that the existing
definition is adequate to describe assembly language code.