Re: error with d28016d16e9a (change to jsonencode.cc)

[John W. Eaton]

On 10/2/20 4:49 AM, Carlo De Falco wrote:

> So, if it is now accepted to require a C++14 compatible compiler we 
> should just document that and change the current configure test which 
> looks for C++11 to look for C++14 instead.
> Otherwise, the use of "auto" for parameters should be abandoned. I'm not 
> following in detail the discussion on this between jwe and Rik but it 
> seems they are planning some way to work around this ...

We might still decide to require C++14 features but I also pushed the 
following changeset which avoids auto in the lambda expressions and 
moves the captured variables from the unwind_action objects to the 
lambda expressions where possible.

  http://hg.savannah.gnu.org/hgweb/octave/rev/445e5ac1f58d

In most cases, the need for "auto" goes away when the variables are 
captured in the lambda expression instead of being passed as arguments 
to the function generated by the lambda expression.  So I meant to make 
that change first, followed by eliminating the remaining uses of "auto" 
in a separate changeset but forgot to split the change before pushing.

I also pushed a separate change to use the [=] capture default 
specification where possible:

  http://hg.savannah.gnu.org/hgweb/octave/rev/1ac5a76ae91d

If we still decide against using the [=] capture default for style 
reasons, it is easy to backout the changeset.

Please let me know if there are still problems with clang after these 
changes.

jwe