[format] Complex requires clauses get formatted wildly

[danakj]

Clang-format produces this:

#include <concepts>

template <class T, class E>
void Result()
    requires(!((std::is_void_v<T> ||
                std::is_trivially_destructible_v<
                    T>)&&std::is_trivially_destructible_v<E>) &&
             !std::is_void_v<T>)
{}

Instead of something like this:

#include <concepts>

template <class T, class E>
void Result()
    requires(!((std::is_void_v<T> || std::is_trivially_destructible_v<T>) &&
               std::is_trivially_destructible_v<E>) &&
             !std::is_void_v<T>)
{}

• It wraps at the template parameter of is_trivially_destructible_v.
• It puts no spaces around the && after it.

See it in action: https://godbolt.org/z/br3hddTbf

[danakj]

Here's a smaller one that goes wrong already, the && loses its spaces around it.

#include <concepts>

template <class T, class E>
void Result()
    requires(!((std::is_void_v<T> || std::is_void_v<T>)&&::std::is_void_v<E>))
{}

Instead of

#include <concepts>

template <class T, class E>
void Result()
    requires(!((std::is_void_v<T> || std::is_void_v<T>) && ::std::is_void_v<E>))
{}

[llvmbot]

@llvm/issue-subscribers-clang-format

[rymiel]

Further reduction of the second case:

void main()
  requires(((A<T>)&&B))
{}

Pertinent debug information:

 M=1 C=1 T=RequiresClause S=1 F=0 B=0 BK=0 P=23 Name=requires L=91 PPK=2 FakeLParens= FakeRParens=0 II=0x55c6c233a178 Text='requires'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=92 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=1 T=Unknown S=0 F=0 B=0 BK=0 P=140 Name=l_paren L=93 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=1 T=Unknown S=0 F=0 B=0 BK=0 P=160 Name=l_paren L=94 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=1 T=Unknown S=0 F=0 B=0 BK=0 P=180 Name=identifier L=95 PPK=2 FakeLParens= FakeRParens=0 II=0x55c6c236d5b0 Text='A'
 M=0 C=0 T=TemplateOpener S=0 F=0 B=0 BK=0 P=90 Name=less L=96 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='<'
 M=0 C=1 T=Unknown S=0 F=0 B=0 BK=0 P=420 Name=identifier L=97 PPK=2 FakeLParens= FakeRParens=0 II=0x55c6c236d5e0 Text='T'
 M=0 C=0 T=TemplateCloser S=0 F=0 B=0 BK=0 P=330 Name=greater L=98 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='>'
 M=0 C=0 T=CastRParen S=0 F=0 B=0 BK=0 P=90 Name=r_paren L=99 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=0 T=UnaryOperator S=0 F=0 B=0 BK=0 P=160 Name=ampamp L=101 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='&&'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=120 Name=identifier L=102 PPK=2 FakeLParens= FakeRParens=0 II=0x55c6c236d610 Text='B'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=63 Name=r_paren L=103 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=43 Name=r_paren L=104 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'

The && is somehow becoming an unary operator, and the parens before it a cast. Note that the redundant outer parens are required for this bug to manifest. (They aren't redundant if you add the ! in front, though)

[rymiel]

Patch: https://reviews.llvm.org/D153641

[Cons-Cat]

Patch: https://reviews.llvm.org/D153641

Thank you so much! This issue has annoyed me for awhile