https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71982
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Trunk gives:
[apinski@xeond2 src]$ ~/upstream-gcc/bin/gcc t.cc
t.cc: In instantiation of ‘bool operator==(const A<T>&, const A<T>&) [with T =
const B]’:
t.cc:25:30: required from here
t.cc:1:34: error: no match for ‘operator==’ (operand types are ‘const B’ and
‘const B’)
1 | #define MACRO(left, right) *left == *right
| ^
t.cc:15:10: note: in expansion of macro ‘MACRO’
15 | return MACRO(_left, _right);
| ^~~~~
t.cc:10:6: note: candidate: ‘template<class T> bool operator==(const A<T>&,
const A<T>&)’
10 | bool operator==(const A<T>& left, const A<T>& right)
| ^~~~~~~~
t.cc:10:6: note: template argument deduction/substitution failed:
t.cc:1:34: note: ‘const B’ is not derived from ‘const A<T>’
1 | #define MACRO(left, right) *left == *right
| ^
t.cc:15:10: note: in expansion of macro ‘MACRO’
15 | return MACRO(_left, _right);
| ^~~~~
t.cc:24:6: note: candidate: ‘bool operator==(C&, C&)’
24 | bool operator==(C& left, C& right) {
| ^~~~~~~~
t.cc:24:20: note: no known conversion for argument 1 from ‘const B’ to ‘C&’
24 | bool operator==(C& left, C& right) {
| ~~~^~~~
Which looks fine.
While GCC 10.1.0 gives:
t.cc: In instantiation of 'bool operator==(const A<T>&, const A<T>&) [with T =
const B]':
t.cc:25:30: required from here
t.cc:1:34: error: no match for 'operator==' (operand types are 'const B' and
'const B')
1 | #define MACRO(left, right) *left == *right
| ~~~~~~^~~~~~~~~
2 |
|
3 | template <typename T>
| ~~~~~~~~~~~~~~~~~~~~~
4 | struct A {
....
This is on the same machine. This happens on both aarch64 and x86_64.
The problem is not the code is being rejected (correctly) but the warning
location includes so much more than it should in previous versions of GCC.
