same name for type as for an array-variable

[piotr5]

obviously it is possible to use same name for type and for variable-name. compiler doesn't mind. however:

delme-nameclash.c: In function 'int main()':
delme-nameclash.c:4:24: error: local external declaration 'name name []' [-fpermissive]
4 | extern struct name name[];
| ^~~~
delme-nameclash.c:1:8: note: does not match previous declaration 'struct name'
1 | struct name {int a,b,c;};
| ^~~~

and you guessed it:

struct name {int a,b,c;};

int main(){
extern struct name name[];
}

as compiler says, -fpermissive will give no errors but merely a warning. however, g++-7.3.0 has no problem compiling it. is this a habit from old times to do such crazy stuff? the keyword "extern" is at fault here, without it and with an explicit array-size it will compile...

so, which behaviour is correct? did g++-7.3.0 break the c++ standard by accepting such a code? or is it a recent addition to the standard? what is the correct way to import a variable that has been named "name" containing an array of type "struct name" as declared in an external header, all outside of my responsibility, in another package or lib?

[AndreyT]

The fact that GCC issues a diagnostic message for this code (which becomes an "error" in -pedantic-errors mode) is apparently a bug in GCC. The code is perfectly valid.

Clang will accept this code without any diagnostics. Note that Clang is not perfectly flawless here either. A code like this

struct name {int a,b,c;};

int main() {
  extern struct name name[];
}

struct name name[42];

triggers an error message from Clang about non-matching declarations. The error message is justified. However, Clang points to the struct name declaration as the conflicting one, while in reality the conflict is between the two array declarations.