Allow in C the more relaxed rules from C++ for assignments with const T *const *

[alejandro-colomar]

Cc: @AaronBallman , @pinskia

GCC bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17654#c11

There is interest in standardizing this in the committee, and it would be interesting to have Clang do this, to show prior art to the committee. I'm also proposing the same exact change to GCC.

The rules have been proven by C++ to just work, so time has proved that these rules work, unless someone points out any issues that I didn't know about with these rules from C++.

No existing code should break, since it's only removing existing diagnostics, and not adding any new ones.

And const correctness still holds after this relaxation of implicit conversions.

[alejandro-colomar]

alx@debian:~/tmp$ cat c.c
int
main(void)
{
	char               *p;
	const char *const  *pp = &p;
}
alx@debian:~/tmp$ gcc -Wall -Wextra -Wno-unused -x c - <c.c
<stdin>: In function ‘main’:
<stdin>:5:27: error: initialization of ‘const char * const*’ from incompatible pointer type ‘char **’ [-Wincompatible-pointer-types]
alx@debian:~/tmp$ gcc -Wall -Wextra -Wno-unused -x c++ - <c.c
alx@debian:~/tmp$ clang -Weverything -Wno-unused -x c - <c.c
<stdin>:5:22: warning: initializing 'const char *const *' with an expression of type 'char **' discards qualifiers in nested pointer types [-Wincompatible-pointer-types-discards-qualifiers]
    5 |         const char *const  *pp = &p;
      |                             ^    ~~
1 warning generated.
alx@debian:~/tmp$ clang -Weverything -Wno-unused -x c++ - <c.c
alx@debian:~/tmp$ 

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: Alejandro Colomar (alejandro-colomar)

Cc: @AaronBallman , @pinskia

GCC bug: <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17654#c11>

There is interest in standardizing this in the committee, and it would be interesting to have Clang do this, to show prior art to the committee. I'm also proposing the same exact change to GCC.

The rules have been proven by C++ to just work, so time has proved that these rules work, unless someone points out any issues that I didn't know about with these rules from C++.

No existing code should break, since it's only removing existing diagnostics, and not adding any new ones.

And const correctness still holds after this relaxation of implicit conversions.

[llvmbot]

@llvm/issue-subscribers-c

Author: Alejandro Colomar (alejandro-colomar)

Cc: @AaronBallman , @pinskia

GCC bug: <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17654#c11>

There is interest in standardizing this in the committee, and it would be interesting to have Clang do this, to show prior art to the committee. I'm also proposing the same exact change to GCC.

The rules have been proven by C++ to just work, so time has proved that these rules work, unless someone points out any issues that I didn't know about with these rules from C++.

No existing code should break, since it's only removing existing diagnostics, and not adding any new ones.

And const correctness still holds after this relaxation of implicit conversions.