[Dwarf-Discuss] DWARF attribute for pass return loc in hidden parameter

Todd Allen todd.allen at ccur.com
Mon May 7 08:12:31 PDT 2012


For our Ada 95 compiler, we had a bunch of different kinds of artificial
parameters, some of which we expected our debugger to understand and be able to
construct or interpret.  We added a DW_AT_artificial_type attribute which could
be present anywhere a DW_AT_artificial attribute was found.  Its value was one
of a bunch of enums for each different kind of artificial parameter.  One of
them was DW_ART_return_block, which was a construct used for returning
dynamically-sized objects and related data.  That was a construct specific to
our compiler, but a similar approach could be of use to you if you have control
over what the compiler generates too.

On Mon, May 07, 2012 at 08:00:33AM -0400, Ron Brender wrote:
> It has been awhile, but I think what we did on VMS for Itanium was
> 
> 1) If the function declaration has a return type, then (of course)
> there is a returned value.
> 2) If the first formal parameter is artificial (DW_AT_artificial
> present) and has no name (DW_AT_name omitted), then it is interpreted
> as the implicit return parameter (whose location is the important
> info). Otherwise, the return location is ABI defined.
> 
> Note, there is no reason to believe that the generated DWARF for a
> given language construct is necessarily the same for every hardware
> architecture.
> 
> Does that help, or am I missing the point?
> 
> Ron
> 
> -------------------------------------------------------------------------------------------
> On Mon, May 7, 2012 at 12:12 AM, Yao Qi <qiyaoltc at gmail.com> wrote:
> > Hi,
> >
> > We encounter a problem in GDB that we are unable to determine the
> > number of parameters a certain function is needed, if the function is
> > a c++ function, and returns value of object.  (See
> > http://sourceware.org/ml/gdb-patches/2012-05/msg00169.html)
> > I'd like to know if DWARF has such attribute to help debugger to
> > understand about the implicit parameter of a certain function.
> >
> > In GDB, when we do a "inferior call", call a function in debuggee from
> > debugger, GDB has to know the parameters of target function, and pass
> > the right parameters in the right place, usually specified by ABI.
> > When we do this in GDB, (s is a String object)
> >
> >  (gdb) print (const char *) (s=s.substr(0,4))
> >
> > GDB has to pass all parameters to `substr', but the problem is
> > `substr' needs different number of parameters on different arch.  On
> > x86, `substr' expects to see four parameters, "return location",
> > "this", "0", and "4", it complies to C++ ABI, however, on some other
> > targets, such as tic6x,SH,m68k, etc, `substr' expects three
> > parameters, "this", "0", and "4".  For the debug information, debugger
> > is unable to know the difference, as below,
> >
> >  <2><233a>: Abbrev Number: 45 (DW_TAG_subprogram)
> >    <233b>   DW_AT_external    : 1
> >    <233c>   DW_AT_name        : (indirect string, offset: 0x1759): substr
> >    <2340>   DW_AT_decl_file   : 9
> >    <2341>   DW_AT_decl_line   : 2004
> >    <2343>   DW_AT_MIPS_linkage_name: (indirect string, offset: 0x297d): _ZNKSs6substrEjj
> >    <2347>   DW_AT_type        : <0x1160>
> >    <234b>   DW_AT_declaration : 1
> >    <234c>   DW_AT_sibling     : <0x2361>
> >  <3><2350>: Abbrev Number: 15 (DW_TAG_formal_parameter)
> >    <2351>   DW_AT_type        : <0x2466>
> >    <2355>   DW_AT_artificial  : 1
> >  <3><2356>: Abbrev Number: 16 (DW_TAG_formal_parameter)
> >    <2357>   DW_AT_type        : <0x3b>
> >  <3><235b>: Abbrev Number: 16 (DW_TAG_formal_parameter)
> >    <235c>   DW_AT_type        : <0x3b>
> >
> > DWARF has DW_TAG_formal_parameter to describe the parameter for a
> > certain function, but looks the "return location in the first implicit
> > parameter" is not covered by DWARF so far, unless I miss something
> > here.  I am wondering that we may add a new
> > DW_AT_return_loc_in_first_implicit_param to DW_TAG_subprogram, so that
> > debugger can pass correct parameters to them.  Thoughts?
> >
> > --
> > Yao Qi
> >
> > _______________________________________________
> > Dwarf-Discuss mailing list
> > Dwarf-Discuss at lists.dwarfstd.org
> > http://lists.dwarfstd.org/listinfo.cgi/dwarf-discuss-dwarfstd.org
> _______________________________________________
> Dwarf-Discuss mailing list
> Dwarf-Discuss at lists.dwarfstd.org
> http://lists.dwarfstd.org/listinfo.cgi/dwarf-discuss-dwarfstd.org

-- 
Todd Allen
Concurrent Computer Corporation



More information about the Dwarf-Discuss mailing list