https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71560
Bug ID: 71560
Summary: union compound literal initializes wrong union field
Product: gcc
Version: 7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: vries at gcc dot gnu.org
Target Milestone: ---
Consider test.c:
...
#include <stdio.h>
union u { int c; float f; };
int
main (void)
{
float f = 2.0;
union u u;
u = (union u){f};
printf ("%f\n", u.f);
return 0;
}
...
This gives the result:
...
$ gcc test.c
$ ./a.out
0.000000
...
Here ( https://gcc.gnu.org/onlinedocs/gcc/Compound-Literals.html ) I read:
...
Compound literals for scalar types and union types are also allowed, but then
the compound literal is equivalent to a cast.
...
So I'd expect the outcome to be the same as for:
...
#include <stdio.h>
union u { int c; float f; };
int
main (void)
{
float f = 2.0;
union u u;
u = (union u)f;
printf ("%f\n", u.f);
return 0;
}
...
and that gives the expected:
...
$ gcc test.c
$ ./a.out
2.000000
...
AFAICT, the problem is introduced by the front-end, which interprets union init
as initializing the int rather than the float field:
...
{
float f = 2.0e+0;
union u u;
float f = 2.0e+0;
union u u;
u = <<< Unknown tree: compound_literal_expr
union u D.2241 = {.c=(int) f}; >>>;
printf ((const char * restrict) "%f\n", (double) u.f);
return 0;
}
return 0;
...
