[clang-format] Misannotation of annotated function names

[HazardyKnusperkeks]

Taken from #69325:
ANNOTATE() void foo(); will be annotated as:

AnnotatedTokens(L=0, P=0, T=5, C=0):
 M=0 C=0 T=Unknown S=1 F=0 B=0 BK=0 P=0 Name=identifier L=8 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7af7110 Text='ANNOTATE'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=9 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=0 T=FunctionAnnotationRParen S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=10 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=1 T=Unknown S=1 F=0 B=0 BK=0 P=23 Name=void L=15 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7abb610 Text='void'
 M=0 C=0 T=TrailingAnnotation S=1 F=0 B=0 BK=0 P=23 Name=identifier L=19 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7af7140 Text='foo'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=20 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=21 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=semi L=22 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=';'

Obviously it shouldn't be TrailingAnnotation, but FunctionDeclarationName, if ANNOTATE is declared as AttributeMacros the annotation is:

AnnotatedTokens(L=0, P=0, T=5, C=0):
 M=0 C=0 T=AttributeMacro S=1 F=0 B=0 BK=0 P=0 Name=identifier L=8 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba5a7f40 Text='ANNOTATE'
 M=0 C=0 T=AttributeLParen S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=9 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=0 T=AttributeRParen S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=10 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=0 T=Unknown S=1 F=0 B=0 BK=0 P=23 Name=void L=15 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba56bd80 Text='void'
 M=0 C=1 T=FunctionDeclarationName S=1 F=0 B=0 BK=0 P=80 Name=identifier L=19 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba5a8138 Text='foo'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=20 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='('
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=21 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')'
 M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=semi L=22 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=';'

Which gets that one right.
There are multiple questions here for me:

1. What is the difference between an attribute and a function declaration? Just reducing it to one breaks 4 tests.
2. Do we want to work an that, or wait until someone has a real issue with it?
3. Should we allow (or even force?) a break between the AttributeRParen and the return type? The code from [clang-format] Incorrect formatting for annotated constrained functions #69325 is not nicely formatted with the AttributeMacros set.

@mydeveloperday @owenca @rymiel

[llvmbot]

@llvm/issue-subscribers-clang-format

Author: Björn Schäpers (HazardyKnusperkeks)

Taken from #69325: `ANNOTATE() void foo();` will be annotated as: ``` AnnotatedTokens(L=0, P=0, T=5, C=0): M=0 C=0 T=Unknown S=1 F=0 B=0 BK=0 P=0 Name=identifier L=8 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7af7110 Text='ANNOTATE' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=9 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='(' M=0 C=0 T=FunctionAnnotationRParen S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=10 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')' M=0 C=1 T=Unknown S=1 F=0 B=0 BK=0 P=23 Name=void L=15 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7abb610 Text='void' M=0 C=0 T=TrailingAnnotation S=1 F=0 B=0 BK=0 P=23 Name=identifier L=19 PPK=2 FakeLParens= FakeRParens=0 II=0x561bf7af7140 Text='foo' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=20 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='(' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=21 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=semi L=22 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=';' ``` Obviously it shouldn't be `TrailingAnnotation`, but `FunctionDeclarationName`, if `ANNOTATE` is declared as `AttributeMacros` the annotation is: ``` AnnotatedTokens(L=0, P=0, T=5, C=0): M=0 C=0 T=AttributeMacro S=1 F=0 B=0 BK=0 P=0 Name=identifier L=8 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba5a7f40 Text='ANNOTATE' M=0 C=0 T=AttributeLParen S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=9 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='(' M=0 C=0 T=AttributeRParen S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=10 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')' M=0 C=0 T=Unknown S=1 F=0 B=0 BK=0 P=23 Name=void L=15 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba56bd80 Text='void' M=0 C=1 T=FunctionDeclarationName S=1 F=0 B=0 BK=0 P=80 Name=identifier L=19 PPK=2 FakeLParens= FakeRParens=0 II=0x564eba5a8138 Text='foo' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=l_paren L=20 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text='(' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=140 Name=r_paren L=21 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=')' M=0 C=0 T=Unknown S=0 F=0 B=0 BK=0 P=23 Name=semi L=22 PPK=2 FakeLParens= FakeRParens=0 II=0x0 Text=';' ``` Which gets that one right. There are multiple questions here for me:

1. What is the difference between an attribute and a function declaration? Just reducing it to one breaks 4 tests.
2. Do we want to work an that, or wait until someone has a real issue with it?
3. Should we allow (or even force?) a break between the AttributeRParen and the return type? The code from #69325 is not nicely formatted with the AttributeMacros set.

@mydeveloperday @owenca @rymiel

[rymiel]

What is the difference between an attribute and a function declaration?

Did you mean function annotation? The first case has a FunctionAnnotationRParen, I assume that's what you're referring to.

[HazardyKnusperkeks]

What is the difference between an attribute and a function declaration?

Did you mean function annotation? The first case has a FunctionAnnotationRParen, I assume that's what you're referring to.

And the attributes have AttributeRParen. I raise the question what is the difference in

1. C++ code generally.
2. In clang-format annotations.

The examples for the function annotations in out tests and in #69325 the macros which get FunctionAnnotationRParens they are always macros which are in fact attributes. One difference is that the arguments to the attribute is not given within [[ ]], but in the following parenthesis.

Are there any function annotations, which are not attributes in reality? Should BreakAfterAttributes affect the annotations?

[owenca]

1. What is the difference between an attribute and a function declaration annotation? Just reducing it to one breaks 4 tests.

I think function annotations are one type of attributes.

2. Do we want to work an that, or wait until someone has a real issue with it?

I would wait.

3. Should we allow (or even force?) a break between the AttributeRParen and the return type? The code from [clang-format] Incorrect formatting for annotated constrained functions #69325 is not nicely formatted with the AttributeMacros set.

We should leave it alone for now and let the user use Macros and BreakAfterAttributes as I suggested in #69325.

[HazardyKnusperkeks]

Then we are done here. :)