add support for non trivial objects about ring-buffer HOT 9 OPEN

Can you elaborate on this? Looking through the code it's not obvious what or where a failure would occur.

from ring-buffer.

Currently insert/removal uses copy constructors everywhere, (i.e. T a = b;) which als means memory leak if T is allocating anything.

from ring-buffer.

As I understand, this should work fine if the author of the class has properly implemented the copy assignment operator. For example, if T is std::shared_ptr, the copy assign would first decrement the usage count of the pointer held in a and then increment the usage count of the pointer held in b.

from ring-buffer.

copy assignment

ah, assignment, not constructor.

It will still leak until given slot is reused for new data, not to mention that the buffer array is uninitialized.

I'll add the assert until figuring out the std::move + rval refs and all of the related "what ifs".

from ring-buffer.

I think the slot would hold onto memory until reuse, but that’s much better than a leak because the buffer size is bounded.

Anyway thanks for the library!

from ring-buffer.

The copy assignment a = b is ok if the copy assignment operator of type T is implemented.
It is responsibility of the maintainer of type T to ensure that current resources is freed before copy resources from others.

Is my understand correct?

from ring-buffer.

Currently the buffer array is uninitialized, so that will be very undefined operation.

This would also make the copy constructors unusable for it's intended purpose.

from ring-buffer.

Currently the buffer array is uninitialized, so that will be very undefined operation.

This would also make the copy constructors unusable for it's intended purpose.

I see. So, if the type can be default constructed, then it will be fine?

from ring-buffer.

As long as all members are also trivially (default) constructible.

from ring-buffer.