Are there any C implementations that have a useless one-bit bitfield `int`?
6.7.2.1p9 from n1570 says:
A member of a structure or union can be of any complete object type other than the modified type .123). In addition, a member can be declared to consist of a specified number of bits (including the sign bit, if any). Such a member is called a bit-field; 124), its width is preceded by a colon.
Do I understand correctly that this means that the only member in struct { int bit:1; }
can be the sign bit?
If this is the case, then it follows that the only values that a bitfield may have for some implementations are 0
and -0
, of which -0
may be indistinguishable from 0
post-save or trap.
Are there any actual implementations where only one value can be assigned to such a bitfield?
source to share
How about gcc 4.9.2?
/* gcc -std=c11 -pedantic-errors -Wextra -Werror=all test.c */
#include <stdio.h>
#include <string.h>
struct foo {
int bit:1;
};
int main() {
struct foo f;
f.bit = 0;
f.bit = 1;
printf("%i\n", f.bit);
return 0;
}
Compilation:
$ gcc -std=c11 -pedantic-errors -Wextra -Werror=all test.c
test.c: In function ‘main’: test.c:12:10: warning: overflow in
implicit constant conversion [-Woverflow]
f.bit = 1;
^
Running it emits:
$ ./a.out
-1
source to share
In 2's complement, all-bit-1 is the maximum negative value. Thus, the sign bit 0
represents 0
and the sign bit 1
represents -1
.
You are describing the representation of a sign value in your question, in which case the values represented will indeed be +0
and -0
. But I don't know of any C ++ implementation that didn't use 2's padding.
source to share