developer notes: updates for C++11 #8177

No description provided.

concept ACK

I'm personally fine with C-style casts for integer types, because for the semantics are simple, unambiguous, and much shorter than using the C++ style.

However, if people feel strongly about this, I'm fine with strictly avoiding them.

I'm fine with C-style casts too for primitive types. No need to change anything there, they're not more magical than anything else.

For object types and such it makes sense to require c++ casts.

ACK on the scoped_ptr to unique_ptr change.

@laanwj it does make searching for casts and points of concern much easier if you use static_cast and the rest of the C++ family of casts. The C-syntax, while more concise in context, often bears a higher intellectual load as to the broader implications of the action in the long run, IMHO.

The extra safety guarantees are also a boon and pretty excellent for what I would consider safety conscious software.

It also makes casts even uglier, as they should be.

it does make searching for casts and points of concern much easier if you use static_cast and the rest of the C++ family of casts.

How often do you actually want to search for all casts?

The extra safety guarantees are also a boon and pretty excellent for what I would consider safety conscious software.

Can you be more specific here? How are casts of primitive types safer using C++ syntax?

I know that pointer casts are safer using C++ syntax and as they are the dangerous kind of casting. I agree with using C++ syntax when casting from const to non const pointers, or one kind of pointer to another, but I don't agree doing so for all cases where primitive types are converted.

I'm also not looking forward to 20 PRs full of 'convert one type of cast to another'. In the time we run this project we've (as far as I know) never had a bug caused by casting. There's no immediate need here. I wrote those developer suggestions all after having found an issue caused by that particular thing. They're not meant for style preferences.

We should mention this one: avoid shadowing vars, and enable -Wshadow #8105 Very sneaky bug introduced by this: #8102 (comment)

Can you be more specific here? How are casts of primitive types safer using C++ syntax?

I know that pointer casts are safer using C++ syntax and as they are the dangerous kind of casting.

Perhaps I mis-indicated that I was referring only to int style casts, the above was a blanket opinion.

In respect to:

I agree with using C++ syntax when casting from const to non const pointers, or one kind of pointer to another, but I don't agree doing so for all cases where primitive types are converted.

I think this is a good compromise. That said, IMHO, primitive conversions can often [but certainly not always] be an indication that the underlying type is over-constrained or just plain wrong. But that problem isn't fixed by this change, it is just made more obvious.

In light of:

I'm also not looking forward to 20 PRs full of 'convert one type of cast to another'

I too think this may be just another nit-pick enabling PR, but, it is certainly a desirable goal... at some point.

Removed the statement on C-style casts.