I'm not sure if this is an actual bug or we are doing something weird, but I've found that if we have a BUILD file with a target which is a macro that returns a value (such as a struct, with the names of a few internally instantiated targets), then a command like buildozer 'print label srcs' //foo will print an error that the rule cannot be found, even if buildozer 'print label srcs' //foo:* shows that the rule exists.

I created a repo which reproduces the issue: https://github.com/mattnworb/buildozer-macro-return-issue

For instance, in //hello-world/BUILD.bazel, there is a target that creates a java_binary via a macro:

macro_java_binary(
    name = "hello-world",
    srcs = ["src/main/java/com/mattnworb/example/Main.java"],
    main_class = "com.mattnworb.example.Main",
)

❯ buildozer 'print label srcs' //hello-world
//hello-world [src/main/java/com/mattnworb/example/Main.java]

but if the macro is made to return something, and that value is assigned to a variable in the BUILD.bazel file, then buildozer can no longer find the rule/target:

foo = macro_java_binary(
    name = "hello-world",
    srcs = ["src/main/java/com/mattnworb/example/Main.java"],
    main_class = "com.mattnworb.example.Main",
)

❯ buildozer 'print label srcs' //hello-world
/Users/mattbrown/code/buildozer-issue-repro/hello-world/BUILD.bazel: error while executing commands [{[print label srcs]}] on target //hello-world: rule 'hello-world' not found

even though buildozer can execute the command against the target in question, when expanding wildcards:

❯ buildozer 'print label srcs' //hello-world:*
//hello-world [src/main/java/com/mattnworb/example/Main.java]

I think a part of the issue is that