Re: [OpenMethod] review starts on 28th of April

From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2025-05-10 17:22:06

AMDG

On 5/8/25 8:13 PM, Jean-Louis Leroy via Boost wrote:
> <snip lots>
 >
>> Switching the default policy based on NDEBUG is dangerous. It's
>> not uncommon to link code build with NDEBUG to code built without it.
> 
> Hmmm yes I see that. I am completely separating the release and debug policies
> in the new design of the policy system. default_policy is a typedef, and, if I
> understand properly, having a typedef aliasing to different types is not an ODR
> violation in itself, but it makes it easier to create one.
> 
> Do you have a better suggestion?
> 

My preference would be to make release the default and
leave debug to be specified explicitly.

>> I don't understand the name vectored_error_handler. How is it vectored?
> 
> I guess it's boomer-speak... We used to call indirect function calls "vectored".
> Do you have a better name?
> 

I guess I can see that. It's just that, in a C++ context,
when I see vector I immediately think std::vector, which
is a completely different meaning of the word.

Some random ideas:
dyn_error_handler
runtime_error_handler
default_error_handler

>> policies/vptr_vector.hpp:
>> 32: Does using namespace policies do anything?
> 
> It imports the names in the current scope? ;-) I do that in the body of some
> functions.

It just looks odd because policies /is/ the current namespace.

In Christ,
Steven Watanabe