The fixes test out, as does the FIXME that's fixed based on the fixes...
Note that the bug causes bogus rejection of any designated initialization
of char array from a string literal, except for the singular case where the
string literal initializer size exactly matches the target char array size
and is not enclosed in optional braces:
typedef struct
C { char id[4]; } C;
C a = {.id = "abc"}; // g++ accepts iff sizeof(C::c) == sizeof("abc")
C b = {.id = {"abc"}}; // g++ rejects valid (gcc accepts)
C c = {.id = "a"}; // g++ rejects valid (gcc accepts)
I'd expect this to be common in C code bases, so the bug would be hit in
any attempt to compile with g++. From the bugzilla comments, it seems that
the following 'workaround' is being used:
C d = {{.id = "a"}}; // g++ accepts invalid (gcc rejects)
which 'works' in this case but is completely borked, consider:
struct name {char first[32], second[32], third[32];};
name DMR {{.first = "Dennis"}, {.third = "Ritchie"}};
Only g++ accepts, ignores the designators, interprets as positional,
and generates correspondingly invalid output:
DMR:
.string "Dennis"
.zero 25
.string "Ritchie"
.zero 24
.zero 32
