[Bug libstdc++/65840] New: type of result of at least some libstdc++ xmethods is different than real operator

[Bug libstdc++/65840] New: type of result of at least some libstdc++ xmethods is different than real operator

[dje at google dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65840

            Bug ID: 65840
           Summary: type of result of at least some libstdc++ xmethods is
                    different than real operator
           Product: gcc
           Version: 4.9.1
            Status: UNCONFIRMED
          Severity: minor
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dje at google dot com

Using libstdc++-v3/testsuite/libstdc++-xmethods/unique_ptr.cc
as a testcase, and modifying it to ensure the * operator is compiled in:

--- gcc-unique_ptr.cc.~1~       2015-04-21 15:17:37.863001716 -0700
+++ gcc-unique_ptr.cc   2015-04-21 15:18:14.666458223 -0700
@@ -30,7 +30,7 @@ main ()
 // { dg-final { note-test *p 10 } }
 // { dg-final { regexp-test p.get() 0x.* } }

-  return 0;  // Mark SPOT
+  return *p;  // Mark SPOT
 }

 // { dg-final { gdb-test SPOT {} 1 } }

bash$ gdb a.out
(gdb) start
(gdb) n # keep next'ing until we get here:
33        return *p;  // Mark SPOT
(gdb) p *p
$1 = 10
(gdb) disable xm
(gdb) p *p
$2 = (int &) @0x403010: 10
(gdb) pt $1
type = int
(gdb) pt $2
type = int &

Note the difference in type of the result of the xmethod vs the real *
operator. IWBN if they had the same type. If this requires additions to gdb we
can do that.

[Bug libstdc++/65840] type of result of at least some libstdc++ xmethods is different than real operator

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65840

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Well does this matter in practice?  In does set "*p = 1" work when xm is
disable and when it is enable?  If the behavior is the same then it does not
matter in practice.

Or does p &*p work both with and without xmethod?

[Bug libstdc++/65840] type of result of at least some libstdc++ xmethods is different than real operator

[dje at google dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65840

--- Comment #2 from Doug Evans <dje at google dot com> ---
I think it should be more than just a matter of working in practice.

A user may get confused by the difference in the type and wonder if time needs
to be spent investigating the difference. Tools shouldn't do this to users.

[Bug libstdc++/65840] type of result of at least some libstdc++ xmethods is different than real operator

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65840

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-04-21
     Ever confirmed|0                           |1