Re: [PATCH] gnu-v2-abi.c: Add casts

[Pedro Alves <palves at redhat dot com>]

On 10/28/2015 08:50 PM, Simon Marchi wrote:
> On 28 October 2015 at 14:46, Pedro Alves <palves@redhat.com> wrote:
>> I looked at changing these is_destructor_name/is_constructor_name
>> interfaces in order to detangle the boolean result from the ctor/dtor
>> kind return, but then realized that this design goes all the way down
>> to the libiberty demangler interfaces.  E.g, include/demangle.h:
>>
>>  ~~~
>>  /* Return non-zero iff NAME is the mangled form of a constructor name
>>     in the G++ V3 ABI demangling style.  Specifically, return an `enum
>>     gnu_v3_ctor_kinds' value indicating what kind of constructor
>>     it is.  */
>>  extern enum gnu_v3_ctor_kinds
>>          is_gnu_v3_mangled_ctor (const char *name);
>>
>>
>>  enum gnu_v3_dtor_kinds {
>>    gnu_v3_deleting_dtor = 1,
>>    gnu_v3_complete_object_dtor,
>>    gnu_v3_base_object_dtor,
>>    /* These are not part of the V3 ABI.  Unified destructors are generated
>>       as a speed-for-space optimization when the -fdeclone-ctor-dtor option
>>       is used, and are always internal symbols.  */
>>    gnu_v3_unified_dtor,
>>    gnu_v3_object_dtor_group
>>  };
>>  ~~~
>>
>> libiberty/cp-demangle.c:
>>
>>  ~~~
>>  enum gnu_v3_ctor_kinds
>>  is_gnu_v3_mangled_ctor (const char *name)
>>  {
>>    enum gnu_v3_ctor_kinds ctor_kind;
>>    enum gnu_v3_dtor_kinds dtor_kind;
>>
>>    if (! is_ctor_or_dtor (name, &ctor_kind, &dtor_kind))
>>      return (enum gnu_v3_ctor_kinds) 0;
>>    return ctor_kind;
>>  }
>>  ~~~
>>
>> etc.
>>
>> gdb/ChangeLog:
>> 2015-10-27  Pedro Alves  <palves@redhat.com>
>>
>>         * gnu-v2-abi.c (gnuv2_is_destructor_name)
>>         (gnuv2_is_constructor_name): Add casts.
>> ---
>>  gdb/gnu-v2-abi.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/gdb/gnu-v2-abi.c b/gdb/gnu-v2-abi.c
>> index c508b55..6c2b92a 100644
>> --- a/gdb/gnu-v2-abi.c
>> +++ b/gdb/gnu-v2-abi.c
>> @@ -40,7 +40,7 @@ gnuv2_is_destructor_name (const char *name)
>>        || startswith (name, "__dt__"))
>>      return complete_object_dtor;
>>    else
>> -    return 0;
>> +    return (enum dtor_kinds) 0;
>>  }
>>
>>  static enum ctor_kinds
>> @@ -51,7 +51,7 @@ gnuv2_is_constructor_name (const char *name)
>>        || startswith (name, "__ct__"))
>>      return complete_object_ctor;
>>    else
>> -    return 0;
>> +    return (enum ctor_kinds) 0;
>>  }
>>
>>  static int
>> --
>> 1.9.3
>>
> 
> I am ok with the cast, 

OK, thanks.  I'm holding you to that.  :-)
I pushed the cast version in as is.

> but just to note that we could do another way.
> I think it would be safe to add a "not_a_ctor = 0" enum value to
> ctor_kinds (which is defined in gdb), and return that in
> gnuv2_is_constructor_name.  Meanwhile, the function in libiberty for
> gnu-abi-v3 (is_gnu_v3_mangled_ctor) will still return a casted 0,
> which will map to not_a_ctor.
> 
> Of course, the comment
> 
> /* Kinds of constructors.  All these values are guaranteed to be
>    non-zero.  */
> 
> would need to change.  Idem for destructors.

Yeah.  I'm not a big fan of the "not_really_a_thing" enum values.
Feels like 'is_constructor_name (enum ctor_kind *kind_if_true);'
would be a better API.  And then ctor_kinds is by design compatible with the v3 enum
as defined by libiberty, and libiberty does the same cast.  (I should also
note that I haven't found any place in gdb that actually cares about of
different enum values of is_constructor_name, etc.  Everywhere seems to care
only about the boolean result.)

Thanks,
Pedro Alves