hasAnyArgument() spuriously ignores match

[xb8g2pw0]

hasAnyArgument() spuriously ignores a match with the program:

void foo (void *, void *);

struct S {
};

void f() {
   S s;
   foo(&s, &s);
}

and the query:

m callExpr(hasAnyArgument(hasType(asString("S *"))))
0 matches. 

However, the query:

m callExpr(hasArgument(0,hasType(asString("S *"))))

returns the expected result:

...clang_ast_bug_2.cpp:10:2: note: "root" binds here
 10 |         foo(&s, &s);
      |         ^~~~~~~~~~~
1 match.

I am using clang-query 17.0.2 built, I believe, by Mozilla.

[danix800]

hasArgument() ignores any implicit casts & parenthesis around the argument via an extra Expr::IgnoreParenImpCasts(), which I think is not necessary, while hasAnyArgument() does not ignore anything.

So if you want hasAnyArgument() to work on this example, use m callExpr(hasAnyArgument( ignoringParenImpCasts(hasType(asString("S *"))))), and an extra ignoringParenImpCasts():

m callExpr(hasAnyArgument(ignoringParenImpCasts(hasType(asString("S *")))))

Match #1:

...test.cpp:8:4: note: "root" binds here
   foo(&s, &s);
   ^~~~~~~~~~~
1 match.

hasArgument() might need to be fixed but I'm afraid that would break a lot of clients.

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: None (xb8g2pw0)

`hasAnyArgument()` spuriously ignores a match with the program:

void foo (void *, void *);

struct S {
};

void f() {
   S s;
   foo(&s, &s);
}

and the query:

m callExpr(hasAnyArgument(hasType(asString("S *"))))
0 matches. 

However, the query:

m callExpr(hasArgument(0,hasType(asString("S *"))))

returns the expected result:

...clang_ast_bug_2.cpp:10:2: note: "root" binds here
 10 |         foo(&s, &s);
      |         ^~~~~~~~~~~
1 match.

I am using clang-query 17.0.2 built, I believe, by Mozilla.

[llvmbot]

@llvm/issue-subscribers-clang-query

Author: None (xb8g2pw0)

`hasAnyArgument()` spuriously ignores a match with the program:

void foo (void *, void *);

struct S {
};

void f() {
   S s;
   foo(&s, &s);
}

and the query:

m callExpr(hasAnyArgument(hasType(asString("S *"))))
0 matches. 

However, the query:

m callExpr(hasArgument(0,hasType(asString("S *"))))

returns the expected result:

...clang_ast_bug_2.cpp:10:2: note: "root" binds here
 10 |         foo(&s, &s);
      |         ^~~~~~~~~~~
1 match.

I am using clang-query 17.0.2 built, I believe, by Mozilla.

[xb8g2pw0]

Thanks for the info. Probably this should be considered a documentation bug? https://clang.llvm.org/docs/LibASTMatchersReference.html does not describe this difference between hasArgument() and hasAnyArgument().

[AaronBallman]

I think it's a functionality bug and not a documentation bug. The intent behind hasArgument and hasAnyArgument is that hasArgument cares about the position of the argument and hasAnyArgument does not. I don't think it should have different behavior regarding implicit AST nodes.

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. In the comments of the issue, request for it to be assigned to you.
2. Fix the issue locally.
3. Run the test suite locally. Remember that the subdirectories under test/ create fine-grained testing targets, so you can e.g. use make check-clang-ast to only run Clang's AST tests.
4. Create a Git commit.
5. Run git clang-format HEAD~1 to format your changes.
6. Open a pull request to the upstream repository on GitHub. Detailed instructions can be found in GitHub's documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment in the thread below.

[llvmbot]

@llvm/issue-subscribers-good-first-issue

Author: None (xb8g2pw0)

`hasAnyArgument()` spuriously ignores a match with the program:

void foo (void *, void *);

struct S {
};

void f() {
   S s;
   foo(&s, &s);
}

and the query:

m callExpr(hasAnyArgument(hasType(asString("S *"))))
0 matches. 

However, the query:

m callExpr(hasArgument(0,hasType(asString("S *"))))

returns the expected result:

...clang_ast_bug_2.cpp:10:2: note: "root" binds here
 10 |         foo(&s, &s);
      |         ^~~~~~~~~~~
1 match.

I am using clang-query 17.0.2 built, I believe, by Mozilla.

[AaronBallman]

I'm marking this as a good first issue because I think the correct behavior is to not ignore implicit AST nodes in hasArgument. We have the TK_IgnoreUnlessSpelledInSource traversal kind for when the user wants to silently skip implicit nodes.

Note, this may cause some fallout that makes this not quite a good first issue because we may have clang-tidy checks that rely on ignoring the implicit nodes. So the fix may be somewhat more involved due to needing to also repair problems with clang-tidy checks.

CC @PiotrZSL @carlosgalvezp @njames93 @LegalizeAdulthood @sam-mccall for extra opinions and thoughts.

[LegalizeAdulthood]

I say go for it. If anyone fixing this gets weird fallout in clang-tidy, we can help out.