This is needed so that we can use conjunction and other logical
type traits to workaround issue with older GCC versions (8 and
below), when they run into types that have ambiguous constructor
from `0`, see e.g. #2571.
However, using conjunction and friends in the SFINAE constraint
in the template parameter breaks for C++20 and up, due to the new
comparison operator rewriting rules. With C++20, when the compiler
see `a == b`, it also tries `b == a` and collects overload set
for both of these expressions.
In Catch2, this means that e.g. `REQUIRE( 1 == 2 )` would lead
the compiler to check overloads for both `ExprLhs<int> == int`
and `int == ExprLhs<int>`. Since the overload set and SFINAE
constraints assume that `ExprLhs<T>` is always on the left side,
when the compiler tries to resolve the template parameters, all
hell breaks loose and the compilation fails.
By moving the SFINAE constraints to the return type, the compiler
can discard the switched expression without having to resolve
the complex SFINAE constraints, and thus everything works the
way it is supposed to.
Fixes#2571
This is primarily done to support new `std::*_ordering` types,
but the refactoring also supports any other type with this
property.
The compilation overhead is surprisingly low. Testing it with
clang on a Linux machine, compiling our SelfTest project takes
only 2-3% longer with these changes than it takes otherwise.
Closes#2555
The old code caused warnings to fire under MSVC, and Clang <3.8.
I could not find a GCC version where it worked, but I assume that it
did at some point.
This new code causes all of MSVC, GCC, Clang, in current versions,
to emit signed/unsigned comparison warning in test like this:
```cpp
TEST_CASE() {
int32_t i = -1;
uint32_t j = 1;
REQUIRE(i != j);
}
```
Where previously only MSVC would emit the warning.
Fixes#1880
In the future we can expect many more matchers, so let's give them
a place to live.
Also moved matcher-related internal files to `internal` subfolder.
Ideally we should sort out all of our source code, but that will
have to come later.
This commit also forbids composing lvalues of composed matchers, as
per previous deprecation notice. I do not expect this to be contentious
in practice, because there was a bug in that usage for years, and
nobody complained.
Now that the recommended distribution and usage method is proper
library, users can just avoid including the matcher headers to get
basically the same effect.
