https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88822
Bug ID: 88822
Summary: Strange inconsistency between types of qualified
rvalues.
Product: gcc
Version: 9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: anders.granlund.0 at gmail dot com
Target Milestone: ---
Test case (prog.c):
int main()
{
int * const p1 = 0;
int * _Atomic p2 = 0;
int * const _Atomic p3 = 0;
struct S { int x; } s;
s = &*p1; // Type of &*p1 is int * const according to error message.
s = &*p2; // Type of &*p2 is int * according to error message.
s = &*p3; // Type of &*p3 is int * according to error message.
}
Compilation command line:
gcc prog.c -Wall -Wextra -std=c11 -pedantic-errors
Observations:
The following error messages are outputed:
incompatible types when assigning to type 'struct S' from type 'int *
const'
incompatible types when assigning to type 'struct S' from type 'int *'
incompatible types when assigning to type 'struct S' from type 'int *'
From the error messages it looks like GCC thinks that the type of &*p1 is
int * const and that the types of &*p2 and &*p3 are both int * .
So it seems like if the _Atomic qualifier is used, all qualifiers are
removed, else they are keept.
6.5.3.2/3 describing the special case of &* does not say anything about
droping any qualifiers. However since the result of &* is an rvalue the
intention of the standard was probably to remove all qualifiers always.
