Why is it UB to pass a char array as the argument of placement new?

Question

I think the following code is ubiquitous:

struct A { ... };
char buf[1024];
auto ptr = new (buf) A(...);

However, new (buf) A(...); is UB. https://en.cppreference.com/w/cpp/language/lifetime.html says (emphasis mine):

As a special case, objects can be created in arrays of unsigned char or std::byte (since C++17) (in which case it is said that the array provides storage for the object) if:

the lifetime of the array has begun and not ended

the storage for the new object fits entirely within the array

there is no array object that satisfies these constraints nested within the array.

Therefore, the char array is not considered as a valid storage type.

Why not allow the char array to be the argument of placement new?

As far as I understand it (if I am wrong I like to be corrected): Pre C++17, only the unsigned data types where guaranteed to be aligned in memory char can be signed on some platforms. So in theory a char array could not be aligned causing issues when allocating allocating aligned objects — Pepijn Kramer
– Pepijn Kramer, Commented Jun 13 at 5:46
If void* was not considered a valid storage type, how would you create an object in a memory block allocated by operator new, the return type of which is void*? The quote mentions a "special case", which implies that there are other cases as well. — Daniel Langr
– Daniel Langr, Commented Jun 13 at 6:46
This post is explaining the thing: stackoverflow.com/questions/79484153/… — PiotrNycz
– PiotrNycz, Commented Jun 13 at 9:02
Standard is trying to limit the number of valid types for uninitialized objects in ordere to maximize the extent of strict alaising, and optimizations that follow. char is widely used to keep text data, while usage of unsigned char is not versatile. Since C++20, std::byte has become the practical buffer data type. So in fresh and new code, you shouldn't even rely on unsigned char; Future standards may hopefully deprecate it. — Red.Wave
– Red.Wave, Commented Jun 13 at 16:16

T.C. · Accepted Answer · 2025-06-15 19:41:47Z

1

The example you show isn't UB (if buf is suitably aligned and is big enough). "provides storage" affects whether the creation of the new object ends the lifetime of the array. But at least in the code snippet you show, whether buf's lifetime ends is immaterial.

As to why plain char isn't included in the "provides storage" rules...the committee apparently thought that it was a new thing so that it can have stricter rules. This is also consistent with the special treatment of unsigned char and std::byte for indeterminate and erroneous values.

edited Jun 15 at 19:41

answered Jun 15 at 19:36

T.C.

139k18 gold badges306 silver badges440 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

T.C. Jun 16 at 12:38

No, it's not. The new already produces a pointer to the created A object. Not that you are using ptr in any way for that to matter.

xmllmx Jun 16 at 12:42

You are correct. Still another issue: the char array and A are not the same type, so they are defined in the otherwise subclause that means std::launder is needed. See the sections of Storage Reuse at en.cppreference.com/w/cpp/language/lifetime.html

Collectives™ on Stack Overflow

Why is it UB to pass a char array as the argument of placement new?

1 Answer 1

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related