Issue 030814.1: Separate compilation units
Chris Quenelle started the discussion with a question.
[ A very long thread is skipped here with various questions asked
and answered. I *think* this is the extraction of a possible
document change, or at least a hint of one. -- David Anderson
Contributors to the discussion:
Chris Quenelle
Daniel Jacobowitz
Michael Eager
David Anderson
Ron Brender
]
Chris Quenelle
> It seems that with the addition of the DW_TAG_partial_unit tag, the
> term "compilation unit" becomes a little ambiguous (at least to me).
> Does "compilation unit" still refer to any block of data in the
> .debug_info section, that has CU-header? (I think of these as "CU-blocks"
> What term do I use to describe the connected tree of dies that
> originates inside one DW_TAG_compile_unit and spans several CU-blocks?
> (I think of this as "the information for an object file (.o file)")
Ron Brender
|Unfortunately there is a awkward vocabulary problem here. At the time we
|introduced DW_TAG_partial_unit we debated whether the term "compilation
|unit" should include partial units or not. In favor of being inclusive
|is the fact that something like 90% of the statements throughout that DWARF
|spec that refer to "compilation units" are true of both DW_TAG_compile_unit
|and DW_TAG_partial_unit. If compilation unit was not inclusive, then
|we would need to add "and partial units" in lots of places. So we went
|with the inclusive concept. In retrospect that may have been the wrong
|decision. We should probably revisit that choice at the next opportunity.
|(Note: this is not a substantive issue, "merely" one of presentation.)
>> In this paragraph, it's not clear what "ordinary compilation unit"
>> means.
>
>An "ordinary compilation unit" would be one which contained descriptions
>of all entities from it's source or include files, without using any of
>the space compression techniques in Appendix E.
|The complaint about the undefined phrase and the suggested definition
|are both right on. Perhaps we can work that definition into the document...
PROPOSAL: none at this time
--------------------------------
Closed without changes to standard. If there is a specific proposal,
submission is invited.