[Dwarf-Discuss] imported_unit and reference identity

Roland McGrath roland at redhat.com
Mon Jan 19 20:35:45 PST 2009

> It looks wrong to me.   Precisely because #b20 is ambiguous.
> It does not make sense to make 'foo' a partial unit, independent of its 
> namespace.

I don't disagree.  So then what is E.2.3 talking about?

> Instead make namespace A and B be independent partial units
> with independent  foo in them. Then your #b?? has to be unambiguous.

That has zero benefit vs the original format in this example (actually it's
negative).  It does not consolidate any duplicates.

> What am I missing here?

You seem to be presuming the "conservative answer" I described.  Perhaps
the only thing you are missing is how someone (like me) could ever have
contemplated anything different.  

In fact, until I re-read all this stuff recently I had thought that
imported_unit could only appear as a direct child of compile_unit.  But
that's not what the spec says, and Appendix E says things that wouldn't
make sense if it did.

I may be dense and am certainly pedantic, but "... whose declarations
logically belong at the place of the imported unit entry" (3.1.2) does not
suffice to fully specify the meaning of using imported_unit such that I
find the answers to the questions I've posed incontrovertibly obvious.


More information about the Dwarf-Discuss mailing list