[libc++] Make __config_site modular (#134699)

This patch makes the __config_site header modular, which solves various
problems with non-modular headers. This requires going back to
generating the modulemap file, since we only know how to make
__config_site modular when we're not using the per-target runtime dir.

The patch also adds a test that we support
-Wnon-modular-include-in-module, which warns about non-modular includes
from modules.

---------

Co-authored-by: Konstantin Varlamov <varconst@apple.com>

Author: ldionne

libcxx/docs/Contributing.rst

@@ -116,7 +116,7 @@ sure you don't forget anything:
 - Did you add all new named declarations to the ``std`` module?
 - If you added a header:
 
-  - Did you add it to ``include/module.modulemap``?
+  - Did you add it to ``include/module.modulemap.in``?
   - Did you add it to ``include/CMakeLists.txt``?
   - If it's a public header, did you update ``utils/libcxx/header_information.py``?
 

libcxx/include/CMakeLists.txt

@@ -1027,7 +1027,6 @@ set(files
   mdspan
   memory
   memory_resource
-  module.modulemap
   mutex
   new
   numbers
@@ -1667,8 +1666,16 @@ set(files
 configure_file("__config_site.in" "${LIBCXX_GENERATED_INCLUDE_TARGET_DIR}/__config_site" @ONLY)
 configure_file("${LIBCXX_ASSERTION_HANDLER_FILE}" "${LIBCXX_GENERATED_INCLUDE_DIR}/__assertion_handler" COPYONLY)
 
+# We generate the modulemap file so that we can include __config_site in it. For now, we don't know how to
+# make __config_site modular when per-target runtime directories are used.
+if (NOT LLVM_ENABLE_PER_TARGET_RUNTIME_DIR)
+  set(LIBCXX_CONFIG_SITE_MODULE_ENTRY "textual header \"__config_site\"")
+endif()
+configure_file("module.modulemap.in" "${LIBCXX_GENERATED_INCLUDE_DIR}/module.modulemap" @ONLY)
+
 set(_all_includes "${LIBCXX_GENERATED_INCLUDE_TARGET_DIR}/__config_site"
-                  "${LIBCXX_GENERATED_INCLUDE_DIR}/__assertion_handler")
+                  "${LIBCXX_GENERATED_INCLUDE_DIR}/__assertion_handler"
+                  "${LIBCXX_GENERATED_INCLUDE_DIR}/module.modulemap")
 foreach(f ${files})
   set(src "${CMAKE_CURRENT_SOURCE_DIR}/${f}")
   set(dst "${LIBCXX_GENERATED_INCLUDE_DIR}/${f}")
@@ -1726,6 +1733,12 @@ if (LIBCXX_INSTALL_HEADERS)
     PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
     COMPONENT cxx-headers)
 
+  # Install the generated modulemap file to the generic include dir.
+  install(FILES "${LIBCXX_GENERATED_INCLUDE_DIR}/module.modulemap"
+    DESTINATION "${LIBCXX_INSTALL_INCLUDE_DIR}"
+    PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
+    COMPONENT cxx-headers)
+
   if (NOT CMAKE_CONFIGURATION_TYPES)
     add_custom_target(install-cxx-headers
                       DEPENDS cxx-headers

libcxx/include/module.modulemap renamed to libcxx/include/module.modulemap.in

@@ -1,6 +1,7 @@
 // This module contains headers related to the configuration of the library. These headers
 // are free of any dependency on the rest of libc++.
 module std_config [system] {
+  @LIBCXX_CONFIG_SITE_MODULE_ENTRY@ // generated via CMake
   textual header "__config"
   textual header "__configuration/abi.h"
   textual header "__configuration/availability.h"

libcxx/test/libcxx/Wnon_modular_include_in_module.compile.pass.cpp

@@ -0,0 +1,21 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// REQUIRES: target={{.*}}-apple-{{.*}}
+// UNSUPPORTED: c++03
+
+// This test ensures that libc++ supports being compiled with modules enabled and with
+// -Wnon-modular-include-in-module. This effectively checks that we don't include any
+// non-modular header from the library.
+//
+// Since most underlying platforms are not modularized properly, this test currently only
+// works on Apple platforms.
+
+// ADDITIONAL_COMPILE_FLAGS: -Wnon-modular-include-in-module -Wsystem-headers-in-module=std -fmodules -fcxx-modules
+
+#include <vector>

libcxx/test/libcxx/headers_in_modulemap.sh.py

@@ -4,7 +4,7 @@
 sys.path.append(sys.argv[1])
 from libcxx.header_information import all_headers, libcxx_include
 
-with open(libcxx_include / "module.modulemap") as f:
+with open(libcxx_include / "module.modulemap.in") as f:
     modulemap = f.read()
 
 isHeaderMissing = False

libcxx/test/libcxx/lint/lint_headers.sh.py

@@ -11,7 +11,7 @@
 def exclude_from_consideration(path):
     return (
         path.endswith(".txt")
-        or path.endswith(".modulemap")
+        or path.endswith(".modulemap.in")
         or os.path.basename(path) == "__config"
         or os.path.basename(path) == "__config_site.in"
         or os.path.basename(path) == "libcxx.imp"

libcxx/utils/libcxx/header_information.py

@@ -15,7 +15,7 @@
 def _is_header_file(file):
     """Returns whether the given file is a header file, i.e. not a directory or the modulemap file."""
     return not file.is_dir() and not file.name in [
-        "module.modulemap",
+        "module.modulemap.in",
         "CMakeLists.txt",
         "libcxx.imp",
         "__config_site.in",