[Mlir-commits] [mlir] [mlir][CMake] Modify the behavior of `add_mlir_aggregate ` and add `mlir_c_target_link_libraries`. (PR #141435)

[Fabian Mora]

================
@@ -565,6 +580,26 @@ function(add_mlir_aggregate name)
       "    OBJECTS = ${_local_objects}\n"
       "    DEPS = ${_local_deps}\n\n")
   endforeach()
+  
+  if (MLIR_LINK_MLIR_DYLIB)
+    # This builds a regex with all the libraries contained in `libMLIR.so`.
+    # These libraries will be filtered out from the dependencies added by the
+    # EMBED_LIBS.
+    list(GET mlir_libs 0 _libs_regex)
+    list(REMOVE_AT mlir_libs 0)
+    set(_libs_regex "(${_libs_regex}")
+    foreach (lib ${mlir_libs})
+      string(APPEND _libs_regex "|${lib}")
----------------
fabianmcg wrote:

> why? one can freely use [add_mlir_library](https://github.com/llvm/llvm-project/blob/main/mlir/examples/standalone/lib/Standalone/CMakeLists.txt#L1) or [add_mlir_public_c_library](https://github.com/llvm/llvm-project/blob/main/mlir/examples/standalone/lib/CAPI/CMakeLists.txt#L1) downstream right?

I'd discourage that, but yes it's possible. The issue is that `add_mlir_library` has an explicit contract that every downstream user should follow, they all should add `EXCLUDE_FROM_LIBMLIR`. But I recognize this could be simplified.

> does the _deps need to be a genex? If not you can just do [list/set difference](https://stackoverflow.com/a/59578250)?

Nope, as soon as you have `genex` you need `genex` to do any manipulation.


https://github.com/llvm/llvm-project/pull/141435