Re: [Boost-users] [variant] Possible regression from 1.55 to 1.58

Subject: Re: [Boost-users] [variant] Possible regression from 1.55 to 1.58
From: Mario Lang (mlang_at_[hidden])
Date: 2015-08-11 18:54:29

Tim Song <t.canens.cpp_at_[hidden]> writes:

> On Tue, Aug 11, 2015 at 5:46 AM, Mario Lang <mlang_at_[hidden]> wrote:
>
>> The example is a bit silly, since foreach discards the return value of
>> apply_visitor anyway.  So my solution was to change the visitor to
>> define result_type as void.  This works.  However, I am wondering why
>> delayed_visitor does no longer work in this situation?  Is it because I
>> am trying to return a reference?
>
> I believe this is caused by a bug in the trait used to detect whether
> result_type is present. It used to form a pointer to result_type,
> and pointer to references are ill-formed, so the trait reports that
> result_type doesn't exist, and then the overloads of apply_visitor
> do not SFINAE correctly.

Sounds about right, thanks for the hint!

> The bug has been fixed in
> https://github.com/boostorg/variant/commit/aad58b5fe5b2605bf752c8151ae0539df98223f9

So I guess this would be fixed with 1.59 then?

-- 
CYa,
  ⡍⠁⠗⠊⠕