[Bug c++/99637] New: bit_cast doesn't work with padding bits and it should

[Bug c++/99637] New: bit_cast doesn't work with padding bits and it should

[hanicka at hanicka dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

            Bug ID: 99637
           Summary: bit_cast doesn't work with padding bits and it should
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hanicka at hanicka dot net
  Target Milestone: ---

Standard states in https://eel.is/c++draft/bit.cast#2 "Padding bits of the
result are unspecified."

Current trunk GCC (11.0.1 20210316) gives error in presence of padding bits:

https://compiler-explorer.com/z/MhKMb5

```
struct my_field
{
    // change size to 64 and it will work
    unsigned long long a : 63;
};

constexpr unsigned long long get_value(const auto & val) noexcept {
    static_assert(sizeof(val) == sizeof(unsigned long long));

    return __builtin_bit_cast(unsigned long long, val);
}

constexpr auto f = my_field{0};
constexpr auto v = get_value(f);
```

this is the error message GCC gives, but this shouldn't be a hard error

```
<source>:14:29:   in 'constexpr' expansion of 'get_value<my_field>(f)'
<source>:10:31: error: '__builtin_bit_cast' accessing uninitialized byte at
offset 7
   10 |     return __builtin_bit_cast(unsigned long long, val);
```

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The result in your case doesn't have any padding bits, but the source has and
some value bits of the result need to be initialized from those padding bits.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[hanicka at hanicka dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

--- Comment #2 from Hana Dusíková <hanicka at hanicka dot net> ---
I know this is not an argument but MSVC accepts this code, meanwhile I'm asking
Richard what to do about it.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu.org,
                   |                            |redi at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I'm CCing Jason and Jonathan too.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
I think the relevant sentence is "Each bit of the value representation of the
result is equal to the corresponding bit in the object representation of from."

For one of the bits in the result, there is no corresponding bit in the object
representation of from. So the result cannot be created.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #4)
> So the result cannot be created.

Not during constant evaluation, anyway.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #5)
> (In reply to Jonathan Wakely from comment #4)
> > So the result cannot be created.
> 
> Not during constant evaluation, anyway.

And outside of constant evaluation with undefined behavior with the exception
of bit_cast to unsigned char / std::byte?

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Yes. If the result type is a class type and the padding bits in the input
correspond to unsigned char or std::byte subobjects in the result, that's OK
(because the only parts with indeterminate values are byte-like types, the rest
of the result is properly initialized).

I think GCC is correct here.

[Bug c++/99637] bit_cast doesn't work with padding bits and it should

[m.cencora at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99637

m.cencora at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |m.cencora at gmail dot com

--- Comment #8 from m.cencora at gmail dot com ---
I think that gcc is incorrect here, because the standard says:
"Each bit of the value representation of the result is equal to the
corresponding bit in the object representation of from."

Padding bits are part of object representation, so if padding bits are properly
initialized in source object (which is the case here), the values of padding
bits should become a part of value representation of destination object.

Also when reading the proposal p0476r2 it is clearly stated that bit_cast is
suppose to be typesafe and constexpr-compatible alternative to memcpy (which by
definition copies padding bits as is).