This is the mail archive of the dwarf2@corp.sgi.com mailing list for the dwarf2 project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: DW_AT_base_types


As I recall, the intent of DW_AT_base_types was to permit a debugger to
find the definitions of base types which were _not used_ in the current
compilation unit, so it could support casting to unsigned long (for example)
even if unsigned long was not used in the current CU (and thus, presumably,
no DIE describing unsigned long would be present in the current CU).

If a producer wished to support casting to unused base types, it could
either emit all possible base types as top-level type descriptions in
each CU, or it could produce a CU containing (possibly only) the base
types and use AT_base_types to refer to it.  Base types which were
_used_ in the current CU then could either be present locally, or the
"commonized" copy could be referred to (via FORM_ref_addr).

The fact that no producer has implemented AT_base_types argues that it does
not save sufficient space to be worthwhile, but I don't favor eliminating
it from DWARF 2.1, even so.

-- 
Dave Weatherford                Forte Tools
David.Weatherford@Sun.COM       Sun Microsystems, Inc.
begin:vcard 
n:Weatherford;David
tel;work:650 786-8942
x-mozilla-html:TRUE
org:Sun Microsystems, Inc.;Forte Tools
version:2.1
email;internet:David.Weatherford@Sun.COM
title:<img src="http://xidus.net/~weath/weath-small.gif"><br>Staff Engineer
adr;quoted-printable:;;901 San Antonio Road=0D=0AMS UMPK16-305;Palo Alto;CA;94303;
x-mozilla-cpt:;-8304
fn:David Weatherford
end:vcard

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]