RE: [PATCH v3] gdb: c++ify btrace_target_info

[Andrew Burgess <aburgess@redhat.com>]

"Metzger, Markus T" <markus.t.metzger@intel.com> writes:

> Hello Andrew,
>
>>>>> -  gdb::unique_xmalloc_ptr<btrace_target_info> tinfo
>>>>> -    (XCNEW (btrace_target_info));
>>>>> -  tinfo->ptid = ptid;
>>>>> +  std::unique_ptr<linux_btrace_target_info> tinfo
>>>>> +    { new linux_btrace_target_info { ptid } };
>>>>
>>>>We recently added a gdb::make_unique function, it would make sense to
>>>>you use it here (it will eventually become std::make_unique once we
>>>>migrate to C++ 14).
>>>>
>>>>So this could be written as:
>>>>
>>>>  auto tinfo = gdb::make_unique<linux_btrace_target_info> (ptid);
>>>
>>> I'm not a fan of 'auto'.  So this becomes
>>>
>>>   std::unique_ptr<linux_btrace_target_info> tinfo
>>>     { gdb::make_unique<linux_btrace_target_info> (ptid) };
>>
>>This really doesn't feel like an improvement.
>>
>>I also try to avoid excessive use of auto, so I dislike things like:
>>
>>  auto var = some_function (....);
>>
>>because there's no hint what the type of var actually is (without
>>looking at `some_function`).  But in the case of:
>>
>>  auto var = gdb::make_unique<type> (...);
>>
>>The type of var is right there on the line, so duplicating the type
>>information is just noise.  For me this is a perfect use of auto.
>
> A related question is whether
>
>     auto ptr { gdb::make_unique<type> (...) }
>
> is really any better than
>
>     std::unique_ptr<type> ptr { new type (...) }

Well std::make_unique<> exists, so I think like it or not, that's the
preferred C++ approach.

I thought (though I don't claim to be an expert) that, in general,
direct calls to 'new' were to be avoided in C++ code.

Thanks,
Andrew