Re: dwarflint versus linkage_name attributes

[Mark Wielaard <mjw at redhat dot com>]

On Mon, 2011-04-04 at 22:30 +0200, Dodji Seketeli wrote:
> Roland McGrath <roland@hack.frob.com> writes:
> >> So this was breaking GDB because it wouldn't figure easily the
> >> relationship between the constructor (which name is, e.g, 't') and the
> >> anonymous struct which name is, well, anonymous.  So adding the linkage
> >> name attribute [which value is 't'] was the easier way [from GDB's
> >> standpoint] to help fix the issue.
> >
> > This is so that GDB can figure out how to call the constructor to evaluate
> > an expression that constructs an object of this type by performing an
> > inferior call?
> 
> That [though I am not sure if GDB supports the evaluation of "new t"
> yet], or "break t::t".
> 
> But even without going that far, GDB was crashing while reading the
> symbol table because it was confused by seeing a typedef (named 't')
> which underlying type was anonymous, with a constructor named 't'.

I must admit I am still confused. So lets try an actual example (the one
from the bug report):

    class C {
    public:
      C() {}
      ~C() {}
    };
    typedef struct {
      C m;
    } t;
    typedef t s;
    s v;

Which generates the following dwarf for g++ -gdwarf-3 (latest gcc trunk
version):

 [    7a]    structure_type
             byte_size            (data1) 1
             decl_file            (data1) 1
             decl_line            (data1) 6
             MIPS_linkage_name    (string) "1t"
             sibling              (ref4) [    bb]
 [    85]      member
               name                 (string) "m"
               decl_file            (data1) 1
               decl_line            (data1) 7
               type                 (ref4) [    31]
               data_member_location (data1) 0
 [    8f]      subprogram
               name                 (string) "t"
               artificial           (flag) Yes
               declaration          (flag) Yes
               object_pointer       (ref4) [    9c]
               sibling              (ref4) [    a3]
 [    9c]        formal_parameter
                 type                 (ref4) [   165]
                 artificial           (flag) Yes
 [    a3]      subprogram
               name                 (string) "~t"
               artificial           (flag) Yes
               declaration          (flag) Yes
               object_pointer       (ref4) [    ad]
 [    ad]        formal_parameter
                 type                 (ref4) [   165]
                 artificial           (flag) Yes
 [    b3]        formal_parameter
                 type                 (ref4) [    73]
                 artificial           (flag) Yes
 [    bb]    typedef
             name                 (string) "t"
             decl_file            (data1) 1
             decl_line            (data1) 8
             type                 (ref4) [    7a]
 [    c4]    typedef
             name                 (string) "s"
             decl_file            (data1) 1
             decl_line            (data1) 9
             type                 (ref4) [    bb]

So the first question is, how do I demangle this "1t"? Is that just a
temporary name given to the anonymous struct?

Second question is how is gdb actually using this? Why is the
linkage_name on the structure_type, and not on the member and/or
subprogram children?

Thanks,

Mark