Is the "Type of bit field too small for number of bits" error part of the C++ standard?

Question

The following code generates MSVS Compiler Error C2034 in Visual Studio 2008:

struct TestStruct { unsigned short var1 : 7; unsigned short : 9; bool var2 : 1; bool : 15; // C2034 };

error C2034: 'TestStruct::<alignment member>': type of bit field too small for number of bits

However, the following code compiles successfully, which seems kind of silly, because I'd think the compiler could have just done this automatically:

struct TestStruct { unsigned short var1 : 7; unsigned short : 9; bool var2 : 1; bool : 7; bool : 8; };

However, both code snippets compile on my Linux GCC compiler. Is one compiler more correct than the other, according to the C++ Standard? If so, which, and why?

Typically, MSVS is almost always wrong. Clang 3.4.1 also compiles it successfully. — Rakete1111
– Rakete1111, Commented Sep 9, 2016 at 16:01
Why would you want the compiler to silently rewrite your code for you behind your back? You explicitly asked for a 15-bit field of type bool, which is impossible. How should it know that you actually wanted an 8-bit field of type bool, as opposed to creating a 15-bit field of type long, or whatever? (GCC and ICC warn about this. Clang might, too, if I knew the right switch. So is your question just, why does MSVC treat this as an error?) — Cody Gray
– Cody Gray ♦, Commented Sep 9, 2016 at 16:05
@CodyGray If it's impossible, why does GCC on Linux let me do it? And the type shouldn't matter, they're just alignment pieces, no? — Zeph M
– Zeph M, Commented Sep 9, 2016 at 16:06
@Rakete1111 except in this case g++ also complains. coliru.stacked-crooked.com/a/c95cb2b1e77dbdb5 (its a warning, but still) — Borgleader
– Borgleader, Commented Sep 9, 2016 at 16:07
It is certainly not impossible if the compiler is allowed to make assumptions. Few things are. But it violates the principle of least surprise to have the compiler reinterpreting code that you've written behind your back. — Cody Gray
– Cody Gray ♦, Commented Sep 9, 2016 at 16:07

NathanOliver · Accepted Answer · 2016-09-09 16:54:48Z

4

Yes this is a bug in MSVS. That standard states in [class.bit]/1

[...]The value of the integral constant expression may be larger than the number of bits in the object representation (3.9) of the bit-field’s type; in such cases the extra bits are used as padding bits and do not participate in the value representation (3.9) of the bit-field.[...]

So the compiler should have added extra padding and only let you have a number of bits equal to CHAR_BIT * sizeof(bit_field_underlying_type).

edited Sep 9, 2016 at 16:54

answered Sep 9, 2016 at 16:06

NathanOliver

183k29 gold badges323 silver badges439 bronze badges

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

5 Comments

lcs Over a year ago

I wouldn't really call this a bug. It is MSVS not complying with the standard definitely, but the error message is clearly intended to be generated. So "non-standard behavior" is probably a better description.

NathanOliver Over a year ago

@lcs I would say that not being standard conforming is a bug.

NathanOliver Over a year ago

@SebastianLenartowicz It did catch this when gcc and clang did not ;)

Ben Voigt Over a year ago

@lcs: In /Za mode, non-conformance is a bug. In /Ze mode, it is merely an intentional deviation.

Pokechu22 Over a year ago

I've sent feedback to MSVC about this.

Collectives™ on Stack Overflow

Is the "Type of bit field too small for number of bits" error part of the C++ standard?

1 Answer 1

5 Comments

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Linked

Related