Re: [PATCH] gdb: Use vector::emplace_back

[Pedro Alves <palves@redhat.com>]

On 11/09/2016 02:48 PM, John Baldwin wrote:
> On Wednesday, November 09, 2016 12:55:52 PM Pedro Alves wrote:
>> On 11/09/2016 12:42 PM, Yao Qi wrote:
>>
>>> I know leading underscore is used in some projects, so I want to know
>>> is it a C++ code standard that we use trailing underscore in this case or
>>> it is your personal coding habit.  It is the latter.
>>
>> ...
>>
>>> Since the trailing underscore usage like this is not mentioned in C++
>>> code standard, people are free to use or not to use it.  I don't have
>>> a preference on that.
>>
>> OK.  I may ask a couple gcc people for their preference and see about
>> adding it to the docs.  Each detail in the standard is based on 
>> someone's personal preference that had sufficient following/agreement,
>> after all.  :-)
> 
> If the goal is to support -Wshadow then it would be nice to settle on a style
> so it is consistent across the tree.

It's not really about -Wshadow.  In C++, in order to use a
member initializer, like in: 

  cmdarg (cmdarg_kind type_, char *string_)
    : type (type_), string (string_)
  {}

The parameter names really must be different from the
struct's elements.

This:

  cmdarg (cmdarg_kind type, char *string)
    : this->type (type), this->string (string)
  {}

is not valid C++ and does _not_ compile:

src/gdb/main.c:450:7: error: expected identifier before ‘this’
     : this->type (type_), this->string (string_)
       ^

This instead would work:

  cmdarg (cmdarg_kind type, char *string)
  {
    this->type = type;
    this->string = string;
  }

However, using member initializer lists is a better
default, because there are cases where the above using
assignment wouldn't work or wouldn't be as efficient.  E.g., in
case the element being constructed has a heavy constructor, does
not have a default constructor at all, or doesn't have an
assignment operator.

You can find more here:

 http://stackoverflow.com/questions/926752/why-should-i-prefer-to-use-member-initialization-list

Thanks,
Pedro Alves