Re: [PATCH 2/2] Avoid copying in lookup_name_info

[Pedro Alves <palves@redhat.com>]

On 3/31/20 8:15 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom Tromey <tromey@adacore.com> writes:
> 
> Pedro> And if we do that, do we have any case where we really need the
> Pedro> "std::string &&" overload?
> 
> Tom> Indeed we don't seem to need this.
> 
> Actually, I am wrong, there is a spot in ada-lang.c:
> 
>       lookup_name_info name1 (std::string ("<_ada_") + name + '>',
> 			      symbol_name_match_type::FULL);
> 
> What's concerning is that this compiles even without the && overload,
> presumably by just violating the lifetime requirement of lookup_name_info.
> 
> So, I think leaving the && version in-place is best.

Alternatively, you could delete the && version, like:

  lookup_name_info (std::string &&name, ......) = delete;

I think that would catch that case, making it fail to compile,
forcing us to write:

 std::string str = std::string ("<_ada_") + name + '>';
 lookup_name_info name1 (str, symbol_name_match_type::FULL);

I think I'd slightly prefer that for making lookup_name_info's
contract easier to grok, though I'll leave it to you.

Thanks,
Pedro Alves